release: prepare v0.3.5

- Add terminal size detection with _MIN_WIDTH (80) and _MIN_HEIGHT (24) thresholds
- Add constrained CSS layout: single-column grid, hide graph, show text summary
- Add #constrained-summary widget with textual history summary
- Add on_resize handler and _refresh()-based constrained state detection
- Preserve selected-day context across resize transitions
- Ensure all regions (headline, health, service, quit) remain usable when constrained
- Update DailyBarGraph._is_constrained to accept terminal_width parameter
- Add comprehensive tests for constrained layout behavior

Covers TPH-9 and cross-cutting TPH regression proof.
Closes #81

.gitea/workflows/release.yml

@@ -0,0 +1,230 @@
+# Fenris release workflow — release path on Coolify-hosted Gitea runner.
+# The runner is repository-scoped and executes package build, signing, validation,
+# registry publication, and release attachment. Spec: §5, issue #52
+name: Release
+
+on:
+  push:
+    tags:
+      - 'v*'
+  workflow_dispatch:
+
+# Built-in Gitea token needs write access for release assets and package registry.
+permissions:
+  contents: read
+  releases: write
+  packages: write
+
+jobs:
+  release:
+    runs-on: [self-hosted]
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Validate release tag and notes
+        run: |
+          set -euo pipefail
+          VERSION="$(sed -n 's/^version = "\(.*\)"/\1/p' pyproject.toml)"
+          if [ -z "${VERSION}" ]; then
+            echo "::error::could not determine the project version"
+            exit 1
+          fi
+          if [ "${GITHUB_EVENT_NAME}" != "workflow_dispatch" ]; then
+            EXPECTED_TAG="v${VERSION}"
+            ACTUAL_TAG="${GITHUB_REF#refs/tags/}"
+            if [ "${ACTUAL_TAG}" != "${EXPECTED_TAG}" ]; then
+              echo "::error::tag ${ACTUAL_TAG} does not match ${EXPECTED_TAG}"
+              exit 1
+            fi
+          fi
+          python3 scripts/extract_changelog.py CHANGELOG.md "${VERSION}" \
+            --footer packaging/release-footer.md > "${RUNNER_TEMP}/release-body.md"
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+
+      - name: Install build dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y gnupg2 rpm python3-venv
+          python3 -m venv /tmp/fenris-ci
+          /tmp/fenris-ci/bin/pip install --quiet build
+          echo "/tmp/fenris-ci/bin" >> "$GITHUB_PATH"
+          NFPM_VERSION=2.47.0
+          curl --fail --silent --show-error --location \
+            "https://github.com/goreleaser/nfpm/releases/download/v${NFPM_VERSION}/nfpm_${NFPM_VERSION}_Linux_x86_64.tar.gz" \
+            -o /tmp/nfpm.tar.gz
+          sudo tar -xzf /tmp/nfpm.tar.gz -C /usr/local/bin nfpm
+          nfpm --version
+
+      - name: Build packages
+        run: make package
+
+      - name: Import packaging key
+        env:
+          GPG_PRIVATE_KEY: ${{ secrets.GPG_PRIVATE_KEY }}
+        run: |
+          set -euo pipefail
+          if [ -z "${GPG_PRIVATE_KEY}" ]; then
+            echo "::error::GPG_PRIVATE_KEY repository secret is not configured"
+            exit 1
+          fi
+          printf '%s\n' "${GPG_PRIVATE_KEY}" | gpg --batch --import
+          SECRET_FINGERPRINT="$(gpg --batch --list-secret-keys --with-colons 'packaging@bongbetic.com' | awk -F: '$1 == "fpr" { print $10; exit }')"
+          PUBLIC_FINGERPRINT="$(gpg --batch --show-keys --with-colons packaging/keys/fenris-packaging.asc | awk -F: '$1 == "fpr" { print $10; exit }')"
+          if [ -z "${PUBLIC_FINGERPRINT}" ]; then
+            echo "::error::packaging/keys/fenris-packaging.asc has no OpenPGP key"
+            exit 1
+          fi
+          if [ "${SECRET_FINGERPRINT}" != "${PUBLIC_FINGERPRINT}" ]; then
+            echo "::error::packaging public key does not match imported private key"
+            exit 1
+          fi
+          echo "Packaging key fingerprint verified: ${PUBLIC_FINGERPRINT}"
+
+      - name: Sign RPM payload
+        run: make sign-rpm
+
+      - name: Generate and clearsign SHA256SUMS
+        run: |
+          set -euo pipefail
+          VERSION="$(sed -n 's/^version = "\(.*\)"/\1/p' pyproject.toml)"
+          cd dist
+          sha256sum "fenris_${VERSION}_amd64.deb" \
+            "fenris-${VERSION}-1.x86_64.rpm" > SHA256SUMS
+          gpg --batch --yes --clearsign --local-user packaging@bongbetic.com SHA256SUMS
+
+      - name: Validate signatures and checksums
+        run: |
+          set -euo pipefail
+          VERSION="$(sed -n 's/^version = "\(.*\)"/\1/p' pyproject.toml)"
+          RPM="fenris-${VERSION}-1.x86_64.rpm"
+          RPM_VERIFY="$(rpm -Kv "dist/${RPM}" 2>&1)"
+          printf '%s\n' "${RPM_VERIFY}"
+          printf '%s\n' "${RPM_VERIFY}" | grep -Eiq 'signature.*: *ok'
+          gpg --batch --verify dist/SHA256SUMS.asc
+          (cd dist && sha256sum -c SHA256SUMS)
+
+      - name: Remove packaging key material
+        if: always()
+        run: |
+          set +e
+          FINGERPRINT="$(gpg --batch --list-secret-keys --with-colons 'packaging@bongbetic.com' 2>/dev/null | awk -F: '$1 == "fpr" { print $10; exit }')"
+          if [ -n "${FINGERPRINT}" ]; then
+            gpg --batch --yes --delete-secret-keys "${FINGERPRINT}"
+            gpg --batch --yes --delete-keys "${FINGERPRINT}"
+          fi
+
+      - name: Determine version
+        id: version
+        run: echo "version=$(sed -n 's/^version = "\(.*\)"/\1/p' pyproject.toml)" >> "$GITHUB_OUTPUT"
+
+      - name: Upload deb packages to registry
+        env:
+          GITEA_PUBLISH_TOKEN: ${{ secrets.GITEAPACKAGETOKEN }}
+        run: |
+          set -euo pipefail
+          if [ -z "${GITEA_PUBLISH_TOKEN}" ]; then
+            echo "::error::GITEAPACKAGETOKEN repository secret is not configured"
+            exit 1
+          fi
+          VERSION=${{ steps.version.outputs.version }}
+          DEB="fenris_${VERSION}_amd64.deb"
+          for CODENAME in bookworm jammy noble; do
+            STATUS=$(curl --silent --show-error --user "xavierk:${GITEA_PUBLISH_TOKEN}" -X PUT \
+              -T "dist/${DEB}" -o /dev/null -w '%{http_code}' \
+              "https://git.bongbetic.com/api/packages/xavierk/debian/pool/${CODENAME}/main/upload" || true)
+            case "${STATUS}" in
+              200|201|204) echo "Debian ${CODENAME}: uploaded" ;;
+              409) echo "Debian ${CODENAME}: already exists, kept existing package" ;;
+              *) echo "::error::Debian ${CODENAME} upload failed with HTTP ${STATUS}"; exit 1 ;;
+            esac
+          done
+
+      - name: Upload RPM to registry
+        env:
+          GITEA_PUBLISH_TOKEN: ${{ secrets.GITEAPACKAGETOKEN }}
+        run: |
+          set -euo pipefail
+          VERSION=${{ steps.version.outputs.version }}
+          RPM="fenris-${VERSION}-1.x86_64.rpm"
+          STATUS=$(curl --silent --show-error --user "xavierk:${GITEA_PUBLISH_TOKEN}" -X PUT \
+            -T "dist/${RPM}" -o /dev/null -w '%{http_code}' \
+            "https://git.bongbetic.com/api/packages/xavierk/rpm/fenris/upload" || true)
+          case "${STATUS}" in
+            200|201|204) echo "RPM: uploaded" ;;
+            409) echo "RPM: already exists, kept existing package" ;;
+            *) echo "::error::RPM upload failed with HTTP ${STATUS}"; exit 1 ;;
+          esac
+
+      - name: Create Gitea release
+        env:
+          GITEA_PUBLISH_TOKEN: ${{ secrets.GITEAPACKAGETOKEN }}
+        run: |
+          set -euo pipefail
+          if [ -z "${GITEA_PUBLISH_TOKEN}" ]; then
+            echo "::error::GITEAPACKAGETOKEN repository secret is not configured"
+            exit 1
+          fi
+          VERSION=${{ steps.version.outputs.version }}
+          RELEASE_BODY="${RUNNER_TEMP}/release-body.md"
+          if [ ! -s "${RELEASE_BODY}" ]; then
+            echo "::error::validated release body is missing or empty"
+            exit 1
+          fi
+          EXISTING_RELEASE="${RUNNER_TEMP}/existing-release.json"
+          EXISTING=$(curl --silent --show-error -o "${EXISTING_RELEASE}" -w '%{http_code}' \
+            -H "Authorization: token ${GITEA_PUBLISH_TOKEN}" \
+            "https://git.bongbetic.com/api/v1/repos/xavierk/Fenris/releases/tags/v${VERSION}" || true)
+          case "${EXISTING}" in
+            200)
+              echo "Release v${VERSION} exists; resynchronizing its notes"
+              REQUEST="$(python3 scripts/release_request.py --version "${VERSION}" \
+                --body-file "${RELEASE_BODY}" --existing-release "${EXISTING_RELEASE}")"
+              ;;
+            404)
+              REQUEST="$(python3 scripts/release_request.py --version "${VERSION}" \
+                --body-file "${RELEASE_BODY}")"
+              ;;
+            *)
+              echo "::error::release lookup failed with HTTP ${EXISTING}"
+              exit 1
+              ;;
+          esac
+          METHOD="$(printf '%s' "${REQUEST}" | python3 -c "import json,sys; print(json.load(sys.stdin)['method'])")"
+          RELEASE_PATH="$(printf '%s' "${REQUEST}" | python3 -c "import json,sys; print(json.load(sys.stdin)['path'])")"
+          PAYLOAD="$(printf '%s' "${REQUEST}" | python3 -c "import json,sys; print(json.dumps(json.load(sys.stdin)['payload']))")"
+          curl --fail --silent --show-error -X "${METHOD}" \
+            -H "Authorization: token ${GITEA_PUBLISH_TOKEN}" \
+            -H "Content-Type: application/json" \
+            -d "${PAYLOAD}" \
+            "https://git.bongbetic.com/api/v1/repos/xavierk/Fenris${RELEASE_PATH}"
+
+      - name: Attach artifacts to release
+        env:
+          GITEA_PUBLISH_TOKEN: ${{ secrets.GITEAPACKAGETOKEN }}
+        run: |
+          set -euo pipefail
+          VERSION=${{ steps.version.outputs.version }}
+          # Get release ID for this tag
+          RELEASE_JSON=$(curl --fail --silent --show-error \
+            -H "Authorization: token ${GITEA_PUBLISH_TOKEN}" \
+            "https://git.bongbetic.com/api/v1/repos/xavierk/Fenris/releases/tags/v${VERSION}")
+          RELEASE_ID=$(printf '%s' "${RELEASE_JSON}" \
+            | python3 -c "import sys,json; print(json.load(sys.stdin)['id'])")
+          # Attach deb, rpm, and clearsigned checksums once.
+          for FILE in "dist/fenris_${VERSION}_amd64.deb" \
+                       "dist/fenris-${VERSION}-1.x86_64.rpm" \
+                       "dist/SHA256SUMS.asc"; do
+            ASSET_NAME="${FILE##*/}"
+            if python3 -c 'import json,sys; name=sys.argv[1]; sys.exit(0 if any(a.get("name") == name for a in json.load(sys.stdin).get("assets", [])) else 1)' "${ASSET_NAME}" <<<"${RELEASE_JSON}"; then
+              echo "${ASSET_NAME}: already attached"
+            else
+              curl --fail --silent --show-error -X POST \
+                -H "Authorization: token ${GITEA_PUBLISH_TOKEN}" \
+                -F "attachment=@${FILE}" \
+                "https://git.bongbetic.com/api/v1/repos/xavierk/Fenris/releases/${RELEASE_ID}/assets"
+            fi
+          done

.gitignore

@@ -7,3 +7,14 @@ data/fenris.log
 data/history.jsonl
 data/hourly.jsonl
 plan-dash-changes.md
+
+# Packaging build artifacts
+build/
+dist/
+
+# Local tooling
+graphify-out/
+json
+src/fenris.egg-info/
+.pytest_cache/
+.venv/

CHANGELOG.md

@@ -0,0 +1,28 @@
+# Changelog
+
+<!--
+Maintainers add one user-facing entry to Unreleased with each change. A release
+commit bumps pyproject.toml, renames Unreleased to that bare-semver version and
+an ISO date, then restores an empty Unreleased section; tag that commit. Do not
+backfill releases from before this changelog.
+-->
+
+## [Unreleased]
+
+## [0.3.5] - 2026-09-14
+
+### Added
+
+- Show trustworthy first-use history, qualifying-day progress, and confidence in the dashboard.
+- Add an interactive daily-writes graph with hourly drill-down and evidence-anchored scenario windows.
+- Repair retained observation history safely and keep its retention state visible.
+- Present a unified monitoring status, drive-health context, and Fenris identity in the dashboard.
+- Remember accessible colour and motion preferences and keep the dashboard usable at constrained terminal sizes.
+
+## [0.3.4] - 2026-09-10
+
+### Added
+
+- Add Fenris identity and a polkit authentication notice to the dashboard.
+- Clarify monitoring continuity, deliberate pauses, and quitting in the dashboard and status output.
+- Add per-release notes with installation, verification, and rollback guidance.

CONTEXT.md

@@ -80,6 +80,14 @@ _Avoid_: Uptime, sample count
 One scheduled or on-demand execution of the collector that interrogates the drive and extends the observation history.
 _Avoid_: Poll, daemon tick
 
+**Release**:
+A published version of Fenris: a version tag, its packages in the channel, and its human-readable change notes, all together; a bare tag is not one.
+_Avoid_: Tag, upload, build
+
+**Rollback**:
+Returning to an earlier release by restoring an observation-store snapshot and then installing that release; installing an older package over a newer store is unsupported.
+_Avoid_: Downgrade, version pinning (as a promise)
+
 **Deliberate disable**:
 A monitoring pause made through Fenris's own control path, closing the monitoring period so the paused time is excluded from the usage habit.
 _Avoid_: Manual stop, service stop

Makefile

@@ -0,0 +1,336 @@
+# Fenris Makefile
+# Spec: §10.1-10.6
+
+SHELL := /bin/bash
+PYTHON := python3
+VENV_DIR := /opt/fenris
+VENDOR_DIR := $(VENV_DIR)/vendor
+BIN_DIR := /usr/local/bin
+LIBEXEC_DIR := /usr/libexec/fenris
+UNIT_DIR := /etc/systemd/system
+POLKIT_DIR := /usr/share/polkit-1/actions
+CONF_DIR := /etc/fenris
+DATA_DIR := /var/lib/fenris
+
+# Placement manifest
+MANIFEST := $(DATA_DIR)/manifest.txt
+
+# Legacy history path (IN-4)
+LEGACY_HISTORY := ./data/history.jsonl
+
+.PHONY: help install upgrade uninstall purge update-deps test lint check-python check-smartctl import-legacy stage package-deb package-rpm package generate-test-key sign-rpm checksums clearsign release release-run release-dry-run clean
+
+help:
+	@echo "Fenris NVMe endurance monitor"
+	@echo ""
+	@echo "Targets:"
+	@echo "  install      - Install Fenris (builds wheel, installs to /opt/fenris)"
+	@echo "  upgrade      - Upgrade Fenris (reinstall wheel, sync units)"
+	@echo "  uninstall    - Uninstall Fenris (preserves config and store)"
+	@echo "  purge        - Remove everything including config and store"
+	@echo "  test         - Run tests"
+	@echo "  lint         - Run linter"
+	@echo "  update-deps  - Update dependency pins"
+	@echo "  stage        - Stage packaging tree for nfpm"
+	@echo "  package      - Build deb + rpm packages"
+	@echo "  package-deb  - Build deb package only"
+	@echo "  package-rpm  - Build rpm package only"
+	@echo "  generate-test-key  - Create throwaway GPG key for CI/testing"
+	@echo "  sign-rpm     - Sign RPM payload with packaging key"
+	@echo "  checksums    - Generate SHA256SUMS manifest"
+	@echo "  clearsign    - Clearsign SHA256SUMS with packaging key"
+	@echo "  release      - Full release (build, sign, checksum, print upload steps)"
+	@echo "  release-run  - Execute the full release flow via scripts/release.sh"
+	@echo "  release-dry-run - Dry-run of the release flow (prints commands only)"
+	@echo "  clean        - Remove build artifacts"
+
+# ─── Pre-install gates ──────────────────────────────────────────────────────
+
+check-python:
+	@echo "=== Verifying Python ≥ 3.10 ==="
+	@$(PYTHON) -c "import sys; v=sys.version_info; exit(0 if (v>=(3,10)) else 1)" || { echo "Error: Python 3.10+ required (found $$($(PYTHON) --version 2>&1))"; exit 1; }
+
+check-smartctl:
+	@echo "=== Verifying smartctl ==="
+	@smartctl --version 2>/dev/null | head -1 || { echo "Error: smartctl not found (install smartmontools)"; exit 1; }
+
+# ─── Build ──────────────────────────────────────────────────────────────────
+
+dist/fenris-*.whl: pyproject.toml src/fenris/*.py
+	@mkdir -p dist
+	$(PYTHON) -m pip wheel --no-deps --wheel-dir dist .
+
+# ─── Install ────────────────────────────────────────────────────────────────
+
+install: check-python check-smartctl dist/fenris-*.whl
+	@echo "=== Creating directories ==="
+	@sudo mkdir -p $(LIBEXEC_DIR)
+	@sudo mkdir -p $(CONF_DIR)
+	@sudo mkdir -p $(POLKIT_DIR)
+	
+	@echo "=== Creating data directory (root-written, group-read) ==="
+	@sudo groupadd -f fenris
+	@sudo install -d -o root -g fenris -m 2750 $(DATA_DIR)
+	
+	@echo "=== Installing version-neutral runtime packages ==="
+	@sudo rm -rf $(VENV_DIR)
+	@sudo install -d -m 0755 $(VENDOR_DIR)
+	@sudo $(PYTHON) -m pip install --disable-pip-version-check --no-compile --target $(VENDOR_DIR) -r requirements.txt dist/fenris-*.whl --quiet
+	
+	@echo "=== Installing wrapper ==="
+	@sudo install -m 0755 scripts/fenris $(BIN_DIR)/fenris
+	
+	@echo "=== Installing helpers ==="
+	@sudo install -m 0755 src/fenris/monitor.py $(LIBEXEC_DIR)/fenris-monitor
+	@sudo install -m 0755 src/fenris/collect.py $(LIBEXEC_DIR)/fenris-collect
+	
+	@echo "=== Installing systemd units (dormant — not enabled/started) ==="
+	@sudo install -m 0644 units/fenris-collect.timer $(UNIT_DIR)/
+	@sudo install -m 0644 units/fenris-collect.service $(UNIT_DIR)/
+	@sudo systemctl daemon-reload
+	
+	@echo "=== Installing polkit policy ==="
+	@sudo install -m 0644 polkit/com.bongbetic.fenris.monitor.policy $(POLKIT_DIR)/
+	
+	@echo "=== Recording manifest (IN-2, IN-10) ==="
+	@echo "# Fenris placement manifest — do not edit" | sudo tee $(MANIFEST) > /dev/null
+	@echo "# Generated by: sudo make install" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "# Timestamp: $$(date -u +%%Y-%%m-%%dT%%H:%%M:%%SZ)" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(BIN_DIR)/fenris" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(LIBEXEC_DIR)/fenris-monitor" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(LIBEXEC_DIR)/fenris-collect" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(UNIT_DIR)/fenris-collect.timer" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(UNIT_DIR)/fenris-collect.service" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(POLKIT_DIR)/com.bongbetic.fenris.monitor.policy" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(VENV_DIR)" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(DATA_DIR)" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(CONF_DIR)" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(MANIFEST)" | sudo tee -a $(MANIFEST) > /dev/null
+	
+	@echo "=== Install complete ==="
+	@echo "Units installed but NOT enabled or started (dormant — IN-3)."
+	@echo "To start monitoring: fenris monitor resume"
+	
+	@$(MAKE) --no-print-directory import-legacy
+
+# ─── Legacy import (IN-4, ST-7) ────────────────────────────────────────────
+
+import-legacy:
+	@if [ -f "$(LEGACY_HISTORY)" ]; then \
+		echo "=== Detected legacy history: $(LEGACY_HISTORY) ==="; \
+		echo "Running idempotent import..."; \
+		PYTHONPATH=$(VENDOR_DIR) $(PYTHON) -c "from fenris.legacy import import_legacy_history; from fenris.store import init_store; from pathlib import Path; conn = init_store(Path('$(DATA_DIR)/observations.db')); r = import_legacy_history(conn, Path('$(LEGACY_HISTORY)')); conn.close(); print(f'  Samples imported: {r.get(\"samples_imported\", 0)}'); print(f'  Hours imported: {r.get(\"hours_imported\", 0)}'); print(f'  Malformed lines: {r.get(\"malformed_lines\", 0)}') if not r.get('skipped') else print('  Skipped: already imported')" || echo "  Warning: import failed (non-fatal)"; \
+	else \
+		echo "=== No legacy history found at $(LEGACY_HISTORY) ==="; \
+	fi
+
+# ─── Upgrade ────────────────────────────────────────────────────────────────
+
+upgrade: dist/fenris-*.whl
+	@echo "=== Upgrading Fenris ==="
+	@echo "=== Snapshotting database (IN-6) ==="
+	@sudo cp $(DATA_DIR)/observations.db $(DATA_DIR)/observations.db.bak 2>/dev/null || true
+	
+	@echo "=== Installing new version-neutral runtime packages ==="
+	@sudo rm -rf $(VENDOR_DIR)
+	@sudo install -d -m 0755 $(VENDOR_DIR)
+	@sudo $(PYTHON) -m pip install --disable-pip-version-check --no-compile --target $(VENDOR_DIR) -r requirements.txt dist/fenris-*.whl --quiet
+	
+	@echo "=== Syncing units against manifest ==="
+	@sudo install -m 0644 units/fenris-collect.timer $(UNIT_DIR)/
+	@sudo install -m 0644 units/fenris-collect.service $(UNIT_DIR)/
+	@sudo install -m 0644 polkit/com.bongbetic.fenris.monitor.policy $(POLKIT_DIR)/
+	@sudo install -m 0755 scripts/fenris $(BIN_DIR)/fenris
+	@sudo install -m 0755 src/fenris/monitor.py $(LIBEXEC_DIR)/fenris-monitor
+	@sudo install -m 0755 src/fenris/collect.py $(LIBEXEC_DIR)/fenris-collect
+	@sudo systemctl daemon-reload
+	
+	@echo "=== Updating manifest ==="
+	@echo "# Fenris placement manifest — do not edit" | sudo tee $(MANIFEST) > /dev/null
+	@echo "# Generated by: sudo make upgrade" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "# Timestamp: $$(date -u +%%Y-%%m-%%dT%%H:%%M:%%SZ)" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(BIN_DIR)/fenris" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(LIBEXEC_DIR)/fenris-monitor" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(LIBEXEC_DIR)/fenris-collect" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(UNIT_DIR)/fenris-collect.timer" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(UNIT_DIR)/fenris-collect.service" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(POLKIT_DIR)/com.bongbetic.fenris.monitor.policy" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(VENV_DIR)" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(DATA_DIR)" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(CONF_DIR)" | sudo tee -a $(MANIFEST) > /dev/null
+	@echo "$(MANIFEST)" | sudo tee -a $(MANIFEST) > /dev/null
+	
+	@echo "=== Restarting timer only if contents changed and active (IN-5) ==="
+	@for unit in fenris-collect.timer fenris-collect.service; do \
+		TMPFILE=$$(mktemp); \
+		sudo systemctl cat $$unit > $$TMPFILE 2>/dev/null || true; \
+		if ! diff -q $$TMPFILE $(UNIT_DIR)/$$unit > /dev/null 2>&1; then \
+			if systemctl is-active --quiet $$unit; then \
+				echo "  $$unit changed and active — restarting"; \
+				sudo systemctl restart $$unit; \
+			fi; \
+		fi; \
+		rm -f $$TMPFILE; \
+	done
+	
+	@echo "=== Applying forward-only schema migrations (IN-5, IN-6) ==="
+	@sudo env PYTHONPATH=$(VENDOR_DIR) $(PYTHON) -c "from fenris.store import migrate_to_latest; from pathlib import Path; n = migrate_to_latest(Path('$(DATA_DIR)/observations.db')); print(f'  Migration steps applied: {n}') if n else print('  Schema already current')"
+	
+	@echo "=== Upgrade complete ==="
+
+# ─── Uninstall (IN-7) ──────────────────────────────────────────────────────
+
+uninstall:
+	@echo "=== Uninstalling Fenris ==="
+	
+	@if [ -x $(LIBEXEC_DIR)/fenris-monitor ]; then \
+		echo "=== Performing sanctioned disable (§10.4) ==="; \
+		sudo $(LIBEXEC_DIR)/fenris-monitor disable --now || true; \
+	fi
+	
+	@echo "=== Stopping units ==="
+	@sudo systemctl stop fenris-collect.timer 2>/dev/null || true
+	@sudo systemctl disable fenris-collect.timer 2>/dev/null || true
+	@sudo systemctl daemon-reload
+	
+	@echo "=== Removing installed files (preserving config and store) ==="
+	@-rm -f $(BIN_DIR)/fenris
+	@-rm -f $(LIBEXEC_DIR)/fenris-monitor
+	@-rm -f $(LIBEXEC_DIR)/fenris-collect
+	@-rmdir $(LIBEXEC_DIR) 2>/dev/null || true
+	@-rm -f $(UNIT_DIR)/fenris-collect.timer
+	@-rm -f $(UNIT_DIR)/fenris-collect.service
+	@-rm -f $(POLKIT_DIR)/com.bongbetic.fenris.monitor.policy
+	@sudo rm -rf $(VENV_DIR)
+	@-sudo rm -f $(MANIFEST)
+	
+	@echo "=== Uninstall complete ==="
+	@echo "Config preserved at $(CONF_DIR)"
+	@echo "Store preserved at $(DATA_DIR)"
+
+# ─── Purge ──────────────────────────────────────────────────────────────────
+
+purge: uninstall
+	@echo "=== Purging Fenris ==="
+	@-sudo rm -rf $(CONF_DIR)
+	@-sudo rm -rf $(DATA_DIR)
+	@echo "=== Purge complete ==="
+
+# ─── Test ───────────────────────────────────────────────────────────────────
+
+test:
+	$(PYTHON) -m pytest tests/ -v
+
+# ─── Lint ───────────────────────────────────────────────────────────────────
+
+lint:
+	$(PYTHON) -m ruff check src/ tests/
+
+# ─── Dependencies (IN-8) ───────────────────────────────────────────────────
+
+update-deps:
+	$(PYTHON) -m pip compile pyproject.toml -o requirements.txt
+
+# ─── Packaging (spec §3, §4, §5) ────────────────────────────────────────────
+
+# Version is sourced from pyproject.toml for both formats
+FENRIS_VERSION := $(shell sed -n 's/^version = "\(.*\)"/\1/p' pyproject.toml)
+
+# GPG signing — packaging key UID (spec §4)
+PACKAGING_KEY ?= packaging@bongbetic.com
+
+stage:
+	@echo "=== Staging packaging tree (v$(FENRIS_VERSION)) ==="
+	$(PYTHON) -m pip wheel --no-deps --wheel-dir dist .
+	bash packaging/stage.sh "$(FENRIS_VERSION)"
+
+package-deb: stage
+	@echo "=== Building deb package ==="
+	VERSION="$(FENRIS_VERSION)" nfpm pkg -f packaging/nfpm.yaml -p deb -t dist/
+	@echo "=== deb package built: dist/fenris_$(FENRIS_VERSION)_amd64.deb ==="
+
+package-rpm: stage
+	@echo "=== Building rpm package ==="
+	VERSION="$(FENRIS_VERSION)" nfpm pkg -f packaging/nfpm.yaml -p rpm -t dist/
+	@echo "=== rpm package built: dist/fenris-$(FENRIS_VERSION)-1.x86_64.rpm ==="
+
+package: package-deb package-rpm
+	@echo "=== Both packages built in dist/ ==="
+
+# ─── GPG key management ─────────────────────────────────────────────────────
+
+generate-test-key:
+	@echo "=== Generating throwaway test GPG key ==="
+	@echo "This key is for CI/testing only — never use for real releases."
+	printf '%%no-protection\nKey-Type: RSA\nKey-Length: 3072\nName-Real: Fenris Packaging (TESTING ONLY)\nName-Email: packaging-test@bongbetic.com\nExpire-Date: 0\n%%commit\n' | \
+		gpg --batch --gen-key
+	@echo "=== Test key created. Fingerprint: ==="
+	@gpg --fingerprint packaging-test@bongbetic.com
+
+# ─── Signing ────────────────────────────────────────────────────────────────
+
+sign-rpm: package-rpm
+	@echo "=== Signing RPM payload ==="
+	@rpm --import packaging/keys/fenris-packaging.asc 2>/dev/null || true
+	rpmsign --addsign --define "_gpg_name $(PACKAGING_KEY)" \
+		dist/fenris-$(FENRIS_VERSION)-1.x86_64.rpm
+	@echo "=== RPM signed ==="
+	@rpm -Kv dist/fenris-$(FENRIS_VERSION)-1.x86_64.rpm
+
+checksums: package
+	@echo "=== Generating SHA256SUMS ==="
+	cd dist && sha256sum fenris_$(FENRIS_VERSION)_amd64.deb \
+		fenris-$(FENRIS_VERSION)-1.x86_64.rpm > SHA256SUMS
+	@echo "=== SHA256SUMS written ==="
+	@cat dist/SHA256SUMS
+
+clearsign: checksums
+	@echo "=== Clearsigning SHA256SUMS ==="
+	gpg --batch --yes --clearsign --local-user $(PACKAGING_KEY) \
+		dist/SHA256SUMS
+	@echo "=== SHA256SUMS.asc written ==="
+
+# ─── Release (spec §5) ──────────────────────────────────────────────────────
+
+release: package sign-rpm clearsign
+	@echo ""
+	@echo "=== Release v$(FENRIS_VERSION) ==="
+	@echo ""
+	@echo "Artifacts:"
+	@ls -la dist/fenris_$(FENRIS_VERSION)_amd64.deb \
+		dist/fenris-$(FENRIS_VERSION)-1.x86_64.rpm \
+		dist/SHA256SUMS.asc 2>/dev/null
+	@echo ""
+	@echo "Verify signing (manual):"
+	@echo "  rpm -Kv dist/fenris-$(FENRIS_VERSION)-1.x86_64.rpm"
+	@echo "  gpg --verify dist/SHA256SUMS.asc dist/SHA256SUMS"
+	@echo ""
+	@echo "Upload to registry:"
+	@echo "  curl -X PUT -u user:token -T dist/fenris_$(FENRIS_VERSION)_amd64.deb \\"
+	@echo "    'https://git.bongbetic.com/api/packages/xavierk/debian/pool/bookworm/main/upload'"
+	@echo "  curl -X PUT -u user:token -T dist/fenris_$(FENRIS_VERSION)_amd64.deb \\"
+	@echo "    'https://git.bongbetic.com/api/packages/xavierk/debian/pool/jammy/main/upload'"
+	@echo "  curl -X PUT -u user:token -T dist/fenris_$(FENRIS_VERSION)_amd64.deb \\"
+	@echo "    'https://git.bongbetic.com/api/packages/xavierk/debian/pool/noble/main/upload'"
+	@echo "  curl -X PUT -u user:token -T dist/fenris-$(FENRIS_VERSION)-1.x86_64.rpm \\"
+	@echo "    'https://git.bongbetic.com/api/packages/xavierk/rpm/fenris/upload'"
+	@echo ""
+	@echo "Create Gitea release with notes and attach:"
+	@echo "  dist/fenris_$(FENRIS_VERSION)_amd64.deb"
+	@echo "  dist/fenris-$(FENRIS_VERSION)-1.x86_64.rpm"
+	@echo "  dist/SHA256SUMS.asc"
+	@echo ""
+	@echo "Key ceremony: delete the private key after upload."
+	@echo "  See docs/install/signing-key-ceremony.md"
+
+# ─── Automated release flow (issue #52) ──────────────────────────────────────
+
+release-run:
+	bash scripts/release.sh --publish
+
+release-dry-run:
+	bash scripts/release.sh --dry-run
+
+clean:
+	@echo "=== Cleaning build artifacts ==="
+	rm -rf build/stage dist/fenris-*.deb dist/fenris-*.rpm dist/SHA256SUMS*

README.md

@@ -1,157 +1,239 @@
-<p align="center">
-  <picture>
-    <source srcset="assets/bongbetic-brand/wordmark-light.png" media="(prefers-color-scheme: dark)">
-    <img src="assets/bongbetic-brand/wordmark-dark.png" alt="Bongbetic" width="260">
-  </picture>
-  <br>
-  <sub>crafted with stubborn curiosity by <a href="https://bongbetic.com">Bongbetic</a></sub>
-</p>
+# Fenris 🐺
 
-<p align="center">
-  <img src="assets/bongbetic-brand/b_glyph.svg" width="48" alt="Fenris glyph">
-</p>
+*Observes an NVMe drive's real-world use and translates that history into an understandable endurance outlook.*
 
-<h1 align="center">Fenris 🐺 — Your SSD's Tell-All Diary</h1>
-
-<p align="center">
-  <em>Your NVMe drive has been keeping secrets. Fenris makes it confess — in real time.</em>
-  <br>
-  <em>How much did you write today? How long until it taps out? No fairy dust — just your actual bytes.</em>
-</p>
+Fenris is a persistent TUI monitor backed by a short-lived privileged collector on a systemd timer. It reads SMART data every few minutes, stores compact observation history in SQLite, and recomputes a usage-adjusted theoretical lifespan on every screen render — no fairy dust, just your actual bytes.
 
 ---
 
-Fenris is a tiny, stubborn daemon that eavesdrops on your NVMe drive's SMART gossip, writes it down every few minutes, and serves you a live dashboard that actually means something. Not "vibes" — **real GB written in the last 24 hours, real GB/hour, and a real countdown in hours, days, and years until your drive's endurance runs out**.
+## Requirements
 
-> Think of it as a Fitbit for your SSD. Except it doesn't nag you to drink water.
+- **Python ≥ 3.10** (verified at install time)
+- **smartmontools** (`smartctl` — verified at install time)
+- **systemd** with a polkit agent (the collector runs as root oneshot; elevation is exclusively polkit)
 
-## What it actually does (no hand-waving)
+No other OS packages or Python dependencies beyond [Textual](https://textual.textualize.io/) (pinned in the lockfile).
 
-- **Listens** — polls `smartctl -j` on your NVMe device (default every 5 minutes, you pick).
-- **Remembers** — appends every sample to `data/history.jsonl` and rolls up per-hour totals into `data/hourly.jsonl` (survives restarts, rebuilds itself if you yank the power).
-- **Calculates** — rolling 24-hour window: *exact* bytes written in the last 24h, GB/h, GB/day, implied total TBW from `percentage_used`, remaining TB, and a projected life-remaining breakdown. Warming-up badge until it has 24h of coverage — no fake confidence.
-- **Shows off** — dense, live dashboard with wear-over-time + trailing-24h per-hour bars, sticky header, live countdown, and stale warnings if the daemon dozes off.
+## Install from package (recommended)
 
-## You need
+### Debian / Ubuntu (apt)
 
-- **Python 3.7+**
-- **smartmontools** (`smartctl`)
-- Root-ish access to read NVMe SMART (passwordless `smartctl` or just run with `sudo` — your call)
+The Gitea instance Debian registry signs metadata with its own key. Verify the
+instance key fingerprint (TOFU hardening):
 
-### The sudo dance (one time)
+```text
+Fingerprint: <print after first release — paste beside the curl one-liner>
+```
 
-Fenris runs `sudo -n smartctl ...` so it doesn't get stuck asking for a password mid-nap:
+Add the instance key and repository:
 
 ```bash
-sudo visudo
-# add this line (swap in your username):
-youruser ALL=(root) NOPASSWD: /usr/sbin/smartctl
+sudo mkdir -p /etc/apt/keyrings
+sudo curl -fsSL -o /etc/apt/keyrings/gitea-xavierk.asc \
+  https://git.bongbetic.com/api/packages/xavierk/debian/repository.key
+
+echo "deb [signed-by=/etc/apt/keyrings/gitea-xavierk.asc] \
+  https://git.bongbetic.com/api/packages/xavierk/debian bookworm main" \
+  | sudo tee /etc/apt/sources.list.d/fenris.list
+
+sudo apt update && sudo apt install fenris
 ```
 
-No sudo? Run the whole thing with `sudo` and it'll still behave.
+Replace `bookworm` with your distribution codename (`bookworm`, `jammy`, or
+`noble`).
 
-## Get it running — 30 seconds
+### Fedora / openSUSE Tumbleweed (RPM)
 
-### The cozy way
+Use the Fenris-owned repo file (not Gitea's auto-generated one):
 
 ```bash
-./fenris.sh
-# pick 1) Start monitoring → choose device / interval / port → done
+sudo dnf config-manager --add-repo \
+  https://git.bongbetic.com/xavierk/Fenris/raw/branch/main/packaging/fenris.repo
+
+sudo dnf install fenris
 ```
 
-### The no-nonsense way
+On openSUSE Tumbleweed, add the same standard RPM repository file and install
+with zypper:
 
 ```bash
-python3 fenris.py start                          # defaults: /dev/nvme0, every 300s, port 8420
-python3 fenris.py start --interval 60 --port 9000  # if you're impatient
-python3 fenris.py status                         # "are we live? how's the drive?"
-python3 fenris.py sample                         # one sneaky sample right now
-python3 fenris.py stop                           # tuck it back in
+sudo zypper addrepo --refresh \
+  https://git.bongbetic.com/xavierk/Fenris/raw/branch/main/packaging/fenris.repo fenris
+sudo zypper install fenris
 ```
 
-Dashboard lives at **http://localhost:8420** (or whatever port you chose).
+The repo file sets `gpgcheck=1` against the Fenris packaging key (downloaded
+from the raw URL in `gpgkey`) and `repo_gpgcheck=0` (metadata check left to
+TLS).
 
-## The menu, demystified
+### Package signature verification
 
-Run `./fenris.sh` and you'll get:
+The RPM payload is signed with the Fenris packaging key (RSA 3072).
+Verification happens automatically via dnf's `gpgcheck=1`. For manual
+verification of downloaded assets:
 
-```
- 1) Start monitoring (background daemon + dashboard)
- 2) Stop monitoring
- 3) Status / current wear stats
- 4) Take one sample right now
- 5) Open dashboard URL
- ---
- h) Help / how this works
- q) Exit (go touch grass)
+```bash
+rpm -Kv fenris-*.x86_64.rpm          # RPM payload signature
+gpg --verify SHA256SUMS.asc SHA256SUMS  # Clearsigned checksum manifest
+sha256sum -c SHA256SUMS                  # Checksum match
 ```
 
-## What Fenris jots down
+The packaging public key is published in-repo — no keyservers. See
+`packaging/keys/fenris-packaging.asc` and
+`docs/install/signing-key-ceremony.md` for key lifecycle details.
 
-| Field | What's the gossip? |
-|-------|---------------------|
-| `percentage_used` | The drive's own wear-o-meter (0–100%) |
-| `bytes_written` / `bytes_read` | Lifetime totals — the receipts |
-| `available_spare` | Spare blocks left (%) |
-| `media_errors` | Uncorrectable boo-boos |
-| `power_on_hours` | How long it's been awake |
-| `temperature_c` | Is it sweating? |
-| `critical_warning` | NVMe's panic flags |
+### Dormant install
 
-Hourly rollups also stash `bytes_written` per hour, `pct_start`/`pct_end`, and temp peaks — so the 24h math stays honest.
+A fresh package install is fully dormant. Units are present but disabled;
+nothing runs. The only opt-in is the sanctioned toggle:
 
-## The dashboard — what's on screen
+```bash
+fenris monitor resume    # enable timer + open first monitoring period
+fenris monitor pause     # close the period, disable timer
+```
 
-- **Hero card: Projected life remaining** — big, friendly `361 d 2 h` (plus `≈ 361 days · ≈ 8666 hours · ≈ 0.99 years`), backed by `~280 GB/day` and `~101 TB left of ~202 TB total` on the test box.
-- **Data written (24h)** — exact GB in the rolling window + coverage (`10.4h of 24h` until warmed up).
-- **Write rate** — GB/h and GB/day, live.
-- **Wear, spare, temp, errors, power-on** — the usual suspects, with progress bars and polite color-coding.
-- **Two charts, side by side:** wear over time + trailing-24h hourly write bars (with a cheeky "now" bar for the current partial hour).
-- **Live plumbing:** polling synced to your interval, ETag-cached, countdown to next sample, warming-up + stale banners, pauses when you hide the tab (saves your battery, you're welcome).
+## Development install (make install)
 
-**API for the tinkerers:** `GET /api/data` · `/api/hourly` · `/api/summary` · `/api/config` · `/api/status` — all JSON, all friendly.
+For contributors building from source:
+
+```bash
+sudo make install
+```
+
+This builds a wheel, installs its locked pure-Python runtime packages into
+`/opt/fenris/vendor`,
+and places helpers, units, and the polkit policy. Units are dormant by default.
+
+```bash
+sudo make upgrade    # re-sync wheel, units, schema
+make uninstall       # removes artifacts, preserves config and store
+make purge           # also removes /etc/fenris and /var/lib/fenris
+```
+
+## Upgrade
+
+### Package upgrade
+
+```bash
+sudo apt update && sudo apt upgrade fenris    # Debian/Ubuntu
+sudo dnf upgrade fenris                       # Fedora
+```
+
+### Development upgrade
+
+```bash
+sudo make upgrade
+```
+
+What it does:
+1. Snapshots `observations.db` to a one-generation backup (`.bak`).
+2. Replaces the locked runtime packages under `/opt/fenris/vendor`.
+3. Syncs units and polkit against the manifest; runs `daemon-reload`.
+4. Restarts the timer **only** if unit contents changed **and** it is active — a running collection run finishes on its mapped interpreter; the next run uses the new code.
+5. Applies forward-only schema migrations (the store directory is never rebuilt; automatic downgrade does not exist).
+
+Rollback: reinstall the previous version and restore `observations.db.bak`.
+
+## Migration from make install
+
+If Fenris was previously installed with `sudo make uninstall` first, then
+installed from the package, existing config, store, and group survive by path
+continuity. Over-installing the package over a `make install` is
+**forbidden** — stale units shadow vendor placement. See
+[docs/install/migrate-from-makeinstall.md](docs/install/migrate-from-makeinstall.md).
+
+## Uninstall and purge
+
+### Package removal
+
+```bash
+sudo apt remove fenris       # preserves config and store
+sudo apt purge fenris        # also removes config and store
+sudo dnf remove fenris       # preserves config and store
+```
+
+### Development removal
+
+```bash
+make uninstall     # removes artifacts, preserves config and observation history
+make purge         # also removes /etc/fenris and /var/lib/fenris
+```
+
+Uninstall performs the sanctioned disable first (`fenris-monitor disable --now`) — an open period closes `user_disabled` — then removes the runtime packages, helpers, units, polkit policy, and wrapper while keeping `/etc/fenris` and the observation store. Reinstalling resumes from the preserved store.
+
+## Cadence drop-ins
+
+The default collection cadence is **5 minutes** (`OnUnitInactiveSec=5min` in the timer unit). To change it, place a systemd drop-in:
+
+```bash
+sudo systemctl edit fenris-collect.timer
+# Add:
+# [Timer]
+# OnUnitInactiveSec=10min
+```
+
+No interval key exists in `/etc/fenris/fenris.conf`. Cadence is a systemd concern, not a Fenris configuration key.
+
+## CLI reference
+
+| Command | Behavior |
+|---|---|
+| `fenris` | Opens the TUI (no arguments). |
+| `fenris status` | Projection facts, enabled/active state, last collect outcome, journal hint on failure or staleness. Never auto-samples. |
+| `fenris sample` | On-demand collection via the privileged helper. Blocks until the run completes. |
+| `fenris monitor pause` | Sanctioned disable — asks for confirmation, then disables the timer and closes the monitoring period. |
+| `fenris monitor resume` | Sanctioned enable — enables the timer and opens a monitoring period. No confirmation. |
+| `fenris baseline set <json>` | CLI-side validation, then polkit-guarded persistence. |
+| `fenris baseline clear` | Remove the endurance baseline. |
+| `fenris import <path>` | Idempotent single-transaction legacy import. |
+| `fenris start` / `stop` / `run` | Rejected with a one-line migration pointer — never aliased. |
+| `fenris --device` | Rejected with a pointer to the configuration file. |
+
+## Reading the dashboard
+
+`fenris` opens the TUI dashboard.
+
+- **Continuity** — the service strip's continuity line (and `fenris status`) reports whether monitoring survives reboots: `monitoring: active in background · persists across reboots`, or `monitoring: does not start on next boot`.
+- **Paused vs. quit** — a full-width `monitoring: paused — deliberate disable` block means collection is stopped (`fenris monitor pause`); resume with `fenris monitor resume`. Pressing `q` only leaves the screen — monitoring keeps running in the background.
+- **Auth banner** — at launch, `privileged actions will prompt for authentication (polkit)` shows once and clears on the first refresh. Privileged actions elevate via polkit; Fenris never asks for sudo.
+
+Per-release notes live on the [releases page](https://git.bongbetic.com/xavierk/Fenris/releases): each entry is the version's `CHANGELOG.md` section — what was added, changed, and fixed — plus standing install and verification instructions.
+
+## Retired menu options
+
+The legacy `fenris.sh` menu script and the `fenris.py` monolith have been removed. Here's where the old options went:
+
+| Legacy option | Successor |
+|---|---|
+| 1) Start monitoring | `fenris monitor resume` |
+| 2) Stop monitoring | `fenris monitor pause` |
+| 3) Status / current wear stats | `fenris status` |
+| 4) Take one sample right now | `fenris sample` |
+| 5) Open dashboard URL | Removed — the HTML dashboard and HTTP server are gone; the TUI is the primary interface. |
+
+## Configuration
+
+`/etc/fenris/fenris.conf` holds exactly one key — the device selector:
+
+```
+device = /dev/disk/by-id/nvme-Samsung_SSD_980_PRO_2TB_S6BENS0Txxxxx
+```
+
+Use a stable `/dev/disk/by-id/` path. Raw `/dev/nvmeX` paths are warned against. The file is re-read every collection run.
 
 ## Where's my stuff?
 
-```
-fenris/
-├── fenris.py                          # the whole show — daemon + server + math
-├── fenris.sh                          # the cozy menu
-├── README.md                          # hi — you're here
-├── assets/bongbetic-brand/            # Bongbetic wordmarks & glyphs (for Gitea + dashboard)
-└── data/
-    ├── history.jsonl                  # raw samples (JSONL, append-only)
-    ├── hourly.jsonl                   # per-hour rollups (auto-rebuilt on restart)
-    ├── fenris.pid                     # daemon PID
-    └── fenris.log                     # daemon chatter
-```
-
-## CLI cheat sheet
-
-```bash
-python3 fenris.py start  [--device /dev/nvme0] [--interval 300] [--port 8420]
-python3 fenris.py stop
-python3 fenris.py status
-python3 fenris.py sample [--device /dev/nvme0]
-python3 fenris.py run    # foreground mode — what `start` spawns internally
-```
-
-## Oops — troubleshooting without the tears
-
-**"smartctl not found"**
-```bash
-sudo apt install smartmontools    # Debian/Ubuntu
-sudo pacman -S smartmontools      # Arch — you already knew
-```
-
-**"needs root" / permission denied**
-Set up the passwordless line above, or just `sudo ./fenris.sh`.
-
-**Dashboard says "stale"**
-Daemon napped or crashed. `python3 fenris.py status` will tell you. Kick it again with `start`.
-
-**Only 10 hours of data and it says "preliminary"?**
-That's honesty, not a bug. It needs 24h of real writes to give a tight estimate. Let it simmer — the number gets sharper every hour.
+| Artifact | Package install | make install |
+|---|---|---|
+| Wrapper | `/usr/bin/fenris` | `/usr/local/bin/fenris` |
+| Helpers | `/usr/libexec/fenris/` | `/usr/libexec/fenris/` |
+| Units | `/usr/lib/systemd/system/` (vendor) | `/etc/systemd/system/` |
+| Polkit policy | `/usr/share/polkit-1/actions/` | `/usr/share/polkit-1/actions/` |
+| sysusers/tmpfiles | `/usr/lib/{sysusers,tmpfiles}.d/fenris.conf` | managed by Makefile |
+| Configuration | `/etc/fenris/fenris.conf` | `/etc/fenris/fenris.conf` |
+| Observation store | `/var/lib/fenris/observations.db` | `/var/lib/fenris/observations.db` |
+| Runtime packages | `/opt/fenris/vendor` | `/opt/fenris/vendor` |
+| Legacy history | — | `./data/history.jsonl` (auto-imported) |
 
 ---
 

docs/adr/0004-install-upgrade-removal-lifecycle.md

@@ -2,7 +2,7 @@
 
 ## Status
 
-Accepted — resolves [Define installation, upgrade, and removal behavior](https://git.bongbetic.com/xavierk/Fenris/issues/9) on the [Wayfinder map](https://git.bongbetic.com/xavierk/Fenris/issues/1).
+Accepted — resolves [Define installation, upgrade, and removal behavior](https://git.bongbetic.com/xavierk/Fenris/issues/9) on the [Wayfinder map](https://git.bongbetic.com/xavierk/Fenris/issues/1). Amended by [ADR 0007](0007-package-delivery-amends-0004.md): package delivery replaces `make install` as primary; layout/ownership and maintainer-script mechanics per 0007. Runtime semantics (dormant install, polkit-only elevation, snapshot + forward-only migration, one-generation rollback) unchanged.
 
 ## Context
 

docs/adr/0007-package-delivery-amends-0004.md

@@ -0,0 +1,34 @@
+# 7. Package delivery: native deb + rpm packages, amending the installation lifecycle
+
+## Status
+
+Accepted — resolves [Task: Compose release spec + ADR amending 0004](https://git.bongbetic.com/xavierk/Fenris/issues/42) on the [Wayfinder map](https://git.bongbetic.com/xavierk/Fenris/issues/33). This ADR **amends [ADR 0004](0004-install-upgrade-removal-lifecycle.md)** on delivery and file ownership only; every runtime semantic of 0004 — dormant install, polkit-only elevation, observation-store snapshot + forward-only migration, one-generation rollback — is inherited verbatim, restated below where the package delivery changes *who* performs it.
+
+## Context
+
+ADR 0004 fixed delivery as `sudo make install` from a source checkout: wheel into a Fenris-owned runtime directory at `/opt/fenris`, a hand-rolled placement manifest, units in `/etc/systemd/system`. The release plan ([map](https://git.bongbetic.com/xavierk/Fenris/issues/33), decisions [Lock channel + toolchain](https://git.bongbetic.com/xavierk/Fenris/issues/38), [Signing + key policy](https://git.bongbetic.com/xavierk/Fenris/issues/39), [Package ownership](https://git.bongbetic.com/xavierk/Fenris/issues/40), [Migration path](https://git.bongbetic.com/xavierk/Fenris/issues/41), [Release cadence](https://git.bongbetic.com/xavierk/Fenris/issues/43)) now ships Fenris as native deb + rpm packages built by nfpm and published to the self-hosted Gitea 1.27.1 package registry, for Debian 12, Ubuntu 22.04/24.04, Fedora 40+, and openSUSE Tumbleweed (x86_64), with locked pure-Python dependencies vendored at `/opt/fenris/vendor`. Packages become the primary delivery; ADR 0004's delivery model demotes to a dev fallback.
+
+The implementation-ready operative contracts live in the [release and packaging specification](../spec/release-packaging.md); this ADR records the decisions and their rationale.
+
+## Decision
+
+Amendments to ADR 0004, section by section:
+
+1. **Delivery (amended).** Packages are primary: one deb per codename pool (`bookworm`, `jammy`, `noble`) and one rpm (group `fenris`, Fedora 40+ and openSUSE Tumbleweed), built by nfpm from a single `packaging/nfpm.yaml` over locked pure-Python runtime packages staged at `/opt/fenris/vendor`, published to the Gitea Debian/RPM registry and installed with `apt`, `dnf`, or `zypper`. `sudo make install` remains as the dev fallback for machines without packages; the two deliveries are mutually exclusive per machine. Version scheme `<pyproject-version>-1`, revision bump on rebuild.
+2. **Layout and manifest (amended).** The hand-rolled manifest model is retired: the dpkg/rpm database **is** the manifest, and nothing like `manifest.txt` ships. Package-owned layout: units in `/usr/lib/systemd/system` (vendor placement; `/etc/systemd/system` is admin-only for drop-ins and enable state); helpers stay in `/usr/libexec/fenris` (exactly `fenris-monitor` and `fenris-collect` — no new polkit-reachable binaries); polkit policy in `/usr/share/polkit-1/actions/`; wrapper at `/usr/bin/fenris` (FHS; `/usr/local/bin` remains `make install`'s). The `fenris` group is declared in `/usr/lib/sysusers.d/fenris.conf` (`g fenris -`) and `/var/lib/fenris` in `/usr/lib/tmpfiles.d/fenris.conf` (`d /var/lib/fenris 2750 root fenris -`), both invoked from the maintainer scripts. The package owns the `/var/lib/fenris` directory only; `observations.db`, WAL sidecars, and `.bak` are never owned and never ghosted — ghost-erase would delete the store, violating 0004 §8.
+3. **Privilege (unchanged).** Root acts through maintainer scripts at install/upgrade/removal time; at runtime, elevation is exclusively polkit, exactly as 0004 §3 and [ADR 0003](0003-service-lifecycle-and-sanctioned-toggle.md) §5 fix it.
+4. **Dormant install (restated for packages).** A fresh package install is fully dormant: postinst/%post performs `systemctl daemon-reload` (plus `systemd-sysusers` and `systemd-tmpfiles --create`) and nothing else — never enable, never preset, never start; no preset file ships. The sanctioned toggle (`fenris monitor resume`) remains the only opt-in.
+5. **Legacy import (narrowed).** Auto-detection of `./data/history.jsonl` is scoped to `make install` only — a package install has no checkout to inspect. `fenris import <path>` remains available as the only import path from packages.
+6. **Upgrade (inherited, maintainer-script mechanics).** Upgrades arrive as packages from the single registry channel. postinst/%post on upgrade: snapshot `observations.db` → one-generation `.bak`, run forward-only schema migrations through the target `python3` with `/opt/fenris/vendor` on its import path (no new binaries), `daemon-reload`, and restart `fenris-collect.timer` only if unit contents changed **and** it is active. `/var/lib/fenris` is never rebuilt; a running oneshot finishes on its old interpreter.
+7. **Rollback (unchanged, plus one hard edge).** One-generation `.bak` semantics are unchanged. Package downgrade is additionally unsupported: forward-only store-version refusal means installing an older package over a newer store fails by design; documented rollback = restore the snapshot, then install the old release.
+8. **Removal (mapped).** deb `remove` ≈ `make uninstall` (conffile and store survive); deb `purge` ≈ `make purge` (plus `.bak` and group cleanup); rpm erase ≈ `make uninstall` (unmodified config removed, modified survives as `.rpmsave`; purge is a documented manual command). prerm/%preun performs the sanctioned disable — `fenris-monitor disable --now`, closing the period `user_disabled` — on remove/erase **only, never on upgrade** (deb prerm upgrade case is a no-op; rpm `%preun` gated on `$1 -eq 0`).
+9. **Conffile semantics (new).** `/etc/fenris/fenris.conf` ships as a placeholder-commented default with no active device selector — deb conffile, rpm `%config(noreplace)`. The device selector is entered by hand (root edits the file), as in both prior deliveries; no configuration verb is added to `fenris-monitor`, and [ADR 0003](0003-service-lifecycle-and-sanctioned-toggle.md) §3's read-and-validate-at-collection-time semantics are untouched. On upgrade, local edits survive as-is; a changed package default lands beside them as `.dpkg-new`/`.rpmnew`.
+10. **Migration from make-install systems (new).** Remove-then-install via runbook only — no migration script, no auto-clean. preinst/%pre aborts with a pointer to the runbook if make-install remnants are detected (`/var/lib/fenris/manifest.txt` or `/etc/systemd/system/fenris-collect.timer`). Store and config survive by path continuity; the migration resets the system to dormant and the user opts back in with `fenris monitor resume`.
+
+## Consequences
+
+- Package installs, upgrades, and removals carry dpkg/rpm-native semantics; nothing in Fenris's own tooling duplicates them.
+- The manifest was 0004's answer to "what did the installer place"; the package database answers it better, and uninstall-keeps-store now holds by package ownership rather than by manifest discipline.
+- `make install` and packages are mutually exclusive per machine; over-install is blocked, not repaired (stale `/etc` units would silently shadow vendor units).
+- Hand-edited configuration remains the model: the device selector is a root-edited file in every delivery, keeping the polkit surface at exactly one binary.
+- Release mechanics — channel, signing, cadence, rollback documentation — are fixed in the [release and packaging specification](../spec/release-packaging.md) and the tickets it cites; this ADR deliberately stops at lifecycle semantics.

docs/install/migrate-from-makeinstall.md

@@ -0,0 +1,87 @@
+# Migrating from make-install to packages
+
+This runbook covers the transition from a `sudo make install` system to the native deb or rpm package. Packages are the primary delivery; `make install` remains as the dev fallback. The two deliveries are **mutually exclusive** per machine.
+
+## Why over-install is forbidden
+
+Installing a package over a make-install system silently breaks things:
+
+- **Stale admin units shadow vendor units.** `make install` places `fenris-collect.timer` and `fenris-collect.service` in `/etc/systemd/system/`. The package installs them in `/usr/lib/systemd/system/` (vendor placement). Systemd loads admin units first — the stale copy takes precedence, and the package update never reaches the running system.
+- **The local wrapper shadows the package wrapper.** `make install` places the `fenris` wrapper at `/usr/local/bin/fenris`. The package places it at `/usr/bin/fenris`. The shell finds `/usr/local/bin` first on PATH — the old checkout-relative wrapper runs instead of the package wrapper.
+
+Neither condition is reversible by reinstalling the package. The only safe path is remove-then-install.
+
+## Pre-migration checklist
+
+1. Confirm no monitoring period is actively running that you want to preserve across the gap:
+   ```
+   fenris status
+   ```
+   The migration resets the system to dormant (see [No-move continuity](#no-move-continuity) below). You opt back in with `fenris monitor resume`.
+
+2. If you have hand-edited configuration at `/etc/fenris/fenris.conf`, note it. The config survives the migration in place (see below).
+
+## Remove step
+
+```
+sudo make uninstall
+```
+
+This performs the **sanctioned disable** (`fenris-monitor disable --now`), closing the current monitoring period as `user_disabled`. It then removes all make-install artifacts: the venv at `/opt/fenris`, the wrapper at `/usr/local/bin/fenris`, the helpers at `/usr/libexec/fenris/`, the units in `/etc/systemd/system/`, and the polkit policy. The placement manifest at `/var/lib/fenris/manifest.txt` is removed.
+
+**What survives the remove:**
+
+- `/var/lib/fenris/observations.db` (and WAL sidecars, `.bak`) — the observation store
+- `/var/lib/fenris/` directory itself — root-written, group-read
+- `/etc/fenris/fenris.conf` — your hand-written configuration
+- The `fenris` system group — created by `groupadd -f` during make-install
+- Journal entries — age out naturally
+
+## Install step
+
+```
+sudo apt install fenris       # Debian/Ubuntu
+sudo dnf install fenris       # Fedora
+```
+
+The package installs into its own layout without touching the surviving store, config, or group.
+
+## No-move continuity
+
+These invariants are verified by the containerized acceptance tests (issue #50):
+
+| Asset | Make-install state | Package post-install | Mechanism |
+|---|---|---|---|
+| `fenris` group | Exists (`groupadd -f`) | Unchanged | `systemd-sysusers` is a no-op when the group already exists |
+| `/var/lib/fenris` directory | Exists (mode 2750, root:fenris) | Unchanged | `systemd-tmpfiles --create` is a no-op when the directory already exists |
+| `observations.db` + sidecars | Present from prior monitoring | Unchanged, never owned by the package | Package owns the directory only; store contents are never ghosted |
+| `/etc/fenris/fenris.conf` | Hand-edited device selector | Survives in place; package default lands as `.dpkg-new` / `.rpmnew` | dpkg conffile / rpm `%config(noreplace)` semantics |
+| Store schema | Version from prior Fenris release | Caught up by the upgrade-path migration | `postinst` / `%post` runs `migrate_to_latest()` on upgrade |
+
+The package detects the make-install system has been removed by the absence of the two markers:
+- `/var/lib/fenris/manifest.txt` (the placement manifest)
+- `/etc/systemd/system/fenris-collect.timer` (pre-manifest make installs)
+
+If either marker exists, the package installation aborts with a pointer to this runbook.
+
+## Reset-to-dormant
+
+`make uninstall`'s sanctioned disable closes the open monitoring period as `user_disabled`. After the package install, the system is dormant — the timer is installed but disabled, nothing is running, no monitoring period is open.
+
+To resume monitoring:
+
+```
+fenris monitor resume
+```
+
+This is the sanctioned opt-in. It enables the timer and opens the first monitoring period in one step. The migration costs at most one short sample gap (the interval between `make uninstall` and `fenris monitor resume`), honestly recorded in the endurance timeline.
+
+## Verification
+
+After migration, confirm the package is correctly installed:
+
+```
+fenris status
+```
+
+The status command should show the dormant state: timer disabled, no active monitoring period, and the observation store intact from the prior make-install system.

docs/install/signing-key-ceremony.md

@@ -0,0 +1,230 @@
+# Signing key ceremony
+
+The Fenris packaging key signs RPM payloads and clearsigns SHA256SUMS manifests.
+This document describes the key's lifecycle: creation, per-release use, rotation,
+and destruction.
+
+## Key specification
+
+| Property | Value |
+|---|---|
+| Algorithm | RSA 3072 |
+| UID | `Fenris Packaging <packaging@bongbetic.com>` |
+| Expiry | 2 years from creation |
+| Hierarchy | Single key — no master/subkey split (single maintainer, manual builds) |
+| Private key storage | Password manager only |
+| Public key storage | `packaging/keys/fenris-packaging.asc` in-repo, release notes, docs |
+| Keyservers | Never — TOFU-over-TLS via raw URL |
+
+## First release: key creation
+
+```bash
+# Generate the dedicated RSA-3072 packaging key
+gpg --batch --gen-key <<EOF
+%no-protection
+Key-Type: RSA
+Key-Length: 3072
+Name-Real: Fenris Packaging
+Name-Email: packaging@bongbetic.com
+Expire-Date: 2y
+%commit
+EOF
+
+# Export the public half — this file is committed to the repo
+gpg --armor --export packaging@bongbetic.com > packaging/keys/fenris-packaging.asc
+
+# Print the fingerprint for docs and release notes
+gpg --fingerprint packaging@bongbetic.com
+```
+
+Save the **private key** to the password manager immediately:
+
+```bash
+gpg --armor --export-secret-keys packaging@bongbetic.com
+```
+
+Then **delete the private key from the local keyring** — it must never persist
+on any build host:
+
+```bash
+gpg --delete-secret-keys packaging@bongbetic.com
+gpg --delete-keys packaging@bongbetic.com
+```
+
+The committed `fenris-packaging.asc` must contain the real public key (replace
+the placeholder comments).
+
+## Per-release signing flow
+
+Each release performs: **import → sign → delete**. The private key is never
+stored on disk longer than the release takes.
+
+### Step 1: Import the private key
+
+Retrieve the private key from the password manager and import it:
+
+```bash
+gpg --import /tmp/packaging-key-private.asc
+rm /f /tmp/packaging-key-private.asc   # Shred if possible
+```
+
+### Step 2: Build and sign packages
+
+The Makefile target `make release` handles signing automatically when the
+key is in the keyring:
+
+```bash
+make release    # builds, signs RPM, clearsigns SHA256SUMS, prints upload steps
+```
+
+Under the hood:
+
+1. `rpmsign --addsign` signs the RPM payload with the packaging key
+   (invoked by `make sign-rpm`).
+2. `sha256sum` generates the checksum manifest.
+3. `gpg --clearsign` produces `SHA256SUMS.asc` with the packaging key.
+
+### Step 3: Delete the private key
+
+Immediately after signing:
+
+```bash
+gpg --delete-secret-keys packaging@bongbetic.com
+gpg --delete-keys packaging@bongbetic.com
+```
+
+Verify the key is gone:
+
+```bash
+gpg --list-keys packaging@bongbetic.com
+# Should produce: gpg: keyblock resource ...: No such file or directory
+```
+
+The entire import → sign → delete cycle should take minutes. The private key
+must never be left in any keyring between releases.
+
+## Key rotation (outline)
+
+When the key approaches expiry, or if it is compromised:
+
+1. **Generate a new key** using the same procedure as first release.
+2. **Publish the new public key** alongside the old one in-repo:
+   ```text
+   packaging/keys/fenris-packaging.asc          # new key (primary)
+   packaging/keys/fenris-packaging-previous.asc  # old key (one cycle)
+   ```
+3. **Sign the next RPM** with the new key.
+4. **Update `fenris.repo`** to list both `gpgkey` URLs (dnf accepts multiple):
+   ```ini
+   gpgkey=https://git.bongbetic.com/xavierk/Fenris/raw/branch/main/packaging/keys/fenris-packaging.asc
+         https://git.bongbetic.com/xavierk/Fenris/raw/branch/main/packaging/keys/fenris-packaging-previous.asc
+   ```
+5. **Drop the old key** from the repo after one release cycle. Delete
+   `fenris-packaging-previous.asc` and revert `gpgkey` to the single URL.
+
+## Verification
+
+Consumers verify the RPM payload signature via dnf (gpgcheck=1 in
+`fenris.repo` points at the published public key). The SHA256SUMS manifest
+verification is manual for downloaded assets:
+
+```bash
+gpg --verify SHA256SUMS.asc SHA256SUMS
+sha256sum -c SHA256SUMS
+```
+
+## One-time live probe
+
+Before the first real release, verify the full registry path end-to-end with a
+throwaway package. This confirms apt/dnf metadata generation, signature
+verification, and consumer setup work as a real consumer would experience them.
+
+### Setup
+
+```bash
+# Create a throwaway package name to avoid polluting fenris metadata
+PROBE_NAME="fenris-regtest"
+PROBE_VERSION="0.0.1"
+```
+
+### Publish
+
+```bash
+# Build a throwaway deb and rpm (use the existing nfpm config with a dummy name)
+# Or use a pre-built package — the probe tests the registry path, not the build
+
+# Upload deb to all codename pools
+for CODENAME in bookworm jammy noble; do
+  curl --fail -X PUT \
+    -u "xavierk:${GITEA_TOKEN}" \
+    -T "dist/${PROBE_NAME}_${PROBE_VERSION}_amd64.deb" \
+    "https://git.bongbetic.com/api/packages/xavierk/debian/pool/${CODENAME}/main/upload"
+done
+
+# Upload rpm
+curl --fail -X PUT \
+  -u "xavierk:${GITEA_TOKEN}" \
+  -T "dist/${PROBE_NAME}-${PROBE_VERSION}-1.x86_64.rpm" \
+  "https://git.bongbetic.com/api/packages/xavierk/rpm/fenris/upload"
+```
+
+### Verify apt metadata (Debian/Ubuntu consumer perspective)
+
+```bash
+# On a Debian/Ubuntu machine:
+sudo mkdir -p /etc/apt/keyrings
+sudo curl -fsSL https://git.bongbetic.com/api/packages/xavierk/debian/repository.key \
+  | sudo gpg --dearmor -o /etc/apt/keyrings/gitea-xavierk.asc
+
+echo "deb [signed-by=/etc/apt/keyrings/gitea-xavierk.asc] https://git.bongbetic.com/api/packages/xavierk/debian bookworm main" \
+  | sudo tee /etc/apt/sources.list.d/fenris.list
+
+sudo apt update
+apt show ${PROBE_NAME}           # metadata present, correct version
+apt install --dry-run ${PROBE_NAME}  # dependency resolution works
+
+# Verify InRelease signature
+apt-key list 2>/dev/null || gpg --no-default-keyring --keyring /etc/apt/keyrings/gitea-xavierk.asc --list-keys
+```
+
+### Verify dnf metadata (Fedora consumer perspective)
+
+```bash
+# On a Fedora machine:
+sudo dnf config-manager --add-repo https://git.bongbetic.com/xavierk/Fenris/raw/branch/main/packaging/fenris.repo
+# Or use Gitea's auto-generated repo for the probe:
+sudo dnf config-manager --add-repo https://git.bongbetic.com/api/packages/xavierk/rpm/fenris.repo
+
+dnf info ${PROBE_NAME}          # metadata present, correct version
+dnf install --assumeno ${PROBE_NAME}  # dependency resolution works
+
+# Verify rpm signature
+rpm -q --scripts ${PROBE_NAME}  # no scripts (throwaway)
+```
+
+### Verify checksums and clearsign
+
+```bash
+# Download from release assets or local build
+gpg --verify SHA256SUMS.asc SHA256SUMS
+sha256sum -c SHA256SUMS
+```
+
+### Cleanup
+
+```bash
+# Delete the throwaway packages from the registry
+for CODENAME in bookworm jammy noble; do
+  curl --fail -X DELETE \
+    -u "xavierk:${GITEA_TOKEN}" \
+    "https://git.bongbetic.com/api/packages/xavierk/debian/pool/${CODENAME}/main/${PROBE_NAME}/${PROBE_VERSION}/amd64"
+done
+
+curl --fail -X DELETE \
+  -u "xavierk:${GITEA_TOKEN}" \
+  "https://git.bongbetic.com/api/packages/xavierk/rpm/fenris/${PROBE_NAME}/${PROBE_VERSION}/x86_64"
+
+# Remove test source list on consumer machines
+sudo rm /etc/apt/sources.list.d/fenris.list
+sudo apt update
+```

docs/research/deb-rpm-toolchain.md

@@ -0,0 +1,239 @@
+# Research: deb + rpm packaging toolchain for bundled-venv builds
+
+Issue: #34 (parent plan: #33) — branch `research/toolchain`
+Date: 2026-09-03 · target: Fenris 0.3.0, x86_64, Debian 12 / Ubuntu 22.04+24.04 / Fedora 40+
+
+## TL;DR
+
+**Recommended: nfpm** with a build script that stages a `--copies` venv at
+`/opt/fenris`. One `nfpm.yaml` is the single source of truth for both formats;
+`nfpm pkg -p deb && nfpm pkg -p rpm` (one invocation per format — `-p` takes a
+single string, verified in `internal/cmd/package.go`). Actively maintained
+(releases v2.47.0, 2026-06-20; repo pushed 2026-08-31). Runner-up: fpm (active,
+v1.18.0 gem 2026-08-26), but its "config" is a long CLI invocation per format —
+the single source of truth degrades into a shell script. dh-virtualenv is
+deb-only and its last upstream release is 2020-10 (effectively dormant);
+rpmbuild spec is rpm-only and cannot share file lists with a deb build without
+external generation.
+
+## Constraint evidence: distro textual is unusable (mostly)
+
+| Distro | python3-textual | Source |
+|---|---|---|
+| Debian 12 (bookworm) | **0.1.13-1** | https://packages.debian.org/bookworm/python3-textual |
+| Ubuntu 22.04 (jammy) | **0.1.13-1** | https://packages.ubuntu.com/jammy/python3-textual |
+| Ubuntu 24.04 (noble) | **0.1.13-1** | https://packages.ubuntu.com/noble/python3-textual |
+| Fedora 40 | 0.48.1 | https://src.fedoraproject.org/rpms/python-textual (f40 spec) |
+| Fedora 41 / 42 / 43 | 0.69.0 / 1.0.0 / 4.0.0 | same spec, f41–f43 branches |
+
+Fenris declares `textual>=0.40.0` (pyproject) but pins `textual==8.2.8`
+(requirements.txt). Debian 12 + Ubuntu 22.04/24.04 are ~1 major era behind even
+the *floor*; Fedora 40 technically meets `>=0.40` but not the pin. Verdict
+unchanged: **vendor deps inside the package for all targets**; per-format
+`depends:` only on `python3 (>= 3.9)`, `smartmontools`, `systemd`.
+
+## Tool-by-tool
+
+### 1. nfpm (goreleaser) — RECOMMENDED
+
+- **Route:** Makefile target builds staging tree → one `nfpm.yaml` →
+  `nfpm package -p deb` + `nfpm package -p rpm`. (Goreleaser release pipeline
+  can wrap both later.)
+- **Shared assets:** version, description, maintainer, `depends`,
+  `contents:` file list, `scripts:` all live once in `nfpm.yaml`; per-format
+  deltas via `overrides: { deb: ..., rpm: ... }` and `packager:`-scoped
+  content entries (https://nfpm.goreleaser.com/configuration/, source
+  `www/content/docs/configuration.md`).
+- **Prerequisites:** single static Go binary (`go install
+  github.com/goreleaser/nfpm/v2/cmd/nfpm@latest`, Homebrew, or release
+  tarball — https://nfpm.goreleaser.com/install/). No toolchain per distro,
+  no root, no containers required (build same tree for both formats).
+- **Venv → file list:** stage with `python3 -m venv --copies staging/opt/fenris
+  && staging/opt/fenris/bin/pip install dist/fenris-*.whl`; map in one entry:
+  `contents: [{ src: staging/opt/fenris/, dst: /opt/fenris, type: tree }]`.
+  Shebangs point at fixed absolute `/opt/fenris/bin/python` → no relocation
+  issues. `--copies` avoids symlink-to-/usr breakage. Config file →
+  `type: config|noreplace` (=%config(noreplace) on rpm, conffile semantics on
+  deb). `/var/lib/fenris` store → `type: ghost` (rpm: owned-but-not-packed;
+  deb: ignored → create in `postinstall` script instead).
+- **Systemd/polkit/libexec:** plain `contents:` entries —
+  `/usr/lib/systemd/system/fenris-collect.{service,timer}` (or
+  `/etc/systemd/system` to match current Makefile), polkit action at
+  `/usr/share/polkit-1/actions/`, helpers under `/usr/libexec/fenris/`.
+  `scripts:` supports `postinstall` (deb maintainer script / rpm scriptlet) —
+  run `systemctl daemon-reload`, create `/var/lib/fenris` root:fenris 2750,
+  group creation.
+- **Upgrade/removal:** deb — dpkg replaces all non-conffile files, conffile
+  prompts/preserves (`.dpkg-new`) per Debian Policy ch-files
+  (https://www.debian.org/doc/debian-policy/ch-files.html); removal keeps
+  conffiles + unowned store; purge cleans. rpm — `rpm -U` replaces,
+  `%config(noreplace)` keeps local edits as `.rpmnew`; only owned dirs are
+  removed on erase (nfpm `type: dir` exists precisely to claim ownership —
+  docs warn not to claim distro-owned dirs).
+- **Maintenance:** very active. goreleaser/nfpm, 2.6k stars, last push
+  2026-08-31, v2.47.0 released 2026-06-20 (GitHub API).
+
+### 2. fpm — viable, weaker single-source-of-truth
+
+- **Route:** staging tree (same as above) then
+  `fpm -s dir -t deb ... staging/=/ ; fpm -s dir -t rpm ...`.
+- **Shared assets:** none declarative — everything is CLI flags
+  (`-n`, `-v`, `--config-files`, `--deb-systemd`, `--directories`,
+  `--after-install`, `--rpm-posttrans`, …). Flag list:
+  https://fpm.readthedocs.io/en/latest/cli-reference.html. The two
+  invocations *will* drift unless wrapped in a Makefile that shares variables;
+  the "single source" is then a shell script, not a checked declarative file.
+  (`--deb-systemd` exists; no rpm-native unit macro — you hand it the unit
+  file plus `--rpm-posttrans` for daemon-reload.)
+- **Prerequisites:** Ruby + gem (`gem install fpm`) or distro package;
+  building rpm side needs `rpmbuild` present for some features.
+- **Venv → file list:** `-s dir` maps a directory into the package verbatim —
+  same staging-tree trick as nfpm. `--config-files /etc/fenris` marks
+  conffiles (deb) / %config (rpm).
+- **Upgrade/removal:** identical downstream semantics to nfpm (native dpkg/rpm
+  behavior); differences are only in how metadata/scripts land in the
+  package.
+- **Maintenance:** active — releases v1.16.0 (2024-12), v1.17.0 (2025-10),
+  v1.18.0 (2026-08-26); gem 1.18.0 on rubygems; ~11.5k stars. But docs are
+  openly "work in progress" (https://fpm.readthedocs.io/en/latest/).
+
+### 3. dh-virtualenv (Spotify) — deb-only, dorms
+
+- **Route:** debhelper add-on: `debian/rules` with
+  `dh $@ --with python-virtualenv --buildsystem=python_distutils`;
+  produces a .deb containing venv at `/opt/venvs/<package>`
+  (`DH_VIRTUALENV_INSTALL_ROOT` overridable, `--builtin-venv` for `python -m
+  venv`). Docs: repo `doc/usage.rst`, `doc/tutorial.rst`
+  (https://github.com/spotify/dh-virtualenv).
+- **Shared assets:** none with rpm — it cannot emit .rpm at all. Would still
+  need a second toolchain for Fedora → fails the criterion outright.
+- **Prerequisites:** `build-essential debhelper devscripts equivs` +
+  `dh-virtualenv` (tutorial.rst); Debian 12 still ships it as
+  `dh-virtualenv 1.2.2-1.3` (https://packages.debian.org/bookworm/dh-virtualenv).
+- **Venv → file list:** automatic — it builds the venv during the debhelper
+  sequence and rewrites shebangs; the .deb owns the whole venv tree. Least
+  manual work of all four, for deb alone.
+- **Upgrade/removal:** standard dpkg; whole venv tree is package-owned, so
+  `apt remove` deletes it cleanly; `--pypi-url`/requirements handled by tool.
+- **Maintenance:** last upstream release **1.2.2, 2020-10-22** (GitHub tag);
+  repo last pushed 2024-04-27, RTD docs 404. Effectively dormant upstream —
+  fine via Debian's own packaging, but risky as strategic dependency.
+- **Bonus fact:** PyPI `dh-virtualenv` project now returns 404 — install only
+  from Debian repo / git.
+
+### 4. rpmbuild spec + vendored venv — rpm-native, no deb
+
+- **Route:** hand-written `fenris.spec`: `%install` stage builds venv into
+  `%{buildroot}/opt/fenris`, `%files` lists it plus units/polkit/libexec,
+  `%ghost %attr(2750,root,fenris) /var/lib/fenris`, `%config(noreplace)` for
+  `/etc/fenris`, `systemd_post/preun` macros for the timer. Reference style:
+  https://docs.fedoraproject.org/en-US/packaging-guidelines/.
+- **Shared assets:** the spec is a second, parallel description of the same
+  file list — nothing is shared with any deb build without generating one
+  side from the other (e.g. generate spec + debian/control from a manifest).
+  Worst single-source-of-truth score.
+- **Prerequisites:** `rpm-build`, mock/koji for cleanroots; Fedora toolchain
+  knowledge; per-distro `Release:`/dist tag handling.
+- **Venv → file list:** `%files` line `%{buildroot}/opt/fenris/...` — venv
+  becomes ordinary payload; shebangs already absolute.
+- **Upgrade/removal:** canonical rpm semantics (same as above) plus real
+  systemd scriptlet macros — the *best-behaved* rpm integration of the four,
+  at the cost of hand-maintained spec.
+- **Maintenance:** rpmbuild itself is maintained forever (part of RPM), but
+  *your* spec is 100% hand-maintained duplication.
+
+## Comparison matrix
+
+| Criterion | nfpm | fpm | dh-virtualenv | rpmbuild spec |
+|---|---|---|---|---|
+| deb + rpm from one config | ✅ one YAML (2 invocations) | ⚠️ flags per invocation | ❌ deb only | ❌ rpm only |
+| File list shared across formats | ✅ `contents:` | ⚠️ per-invocation args | n/a | ❌ |
+| Vendored venv supported | ✅ staging `type: tree` | ✅ `-s dir` | ✅✅ automatic (deb) | ✅ `%files` |
+| conffile / %config(noreplace) | ✅ `type: config\|noreplace` | ✅ `--config-files` | ✅ (debhelper) | ✅ `%config(noreplace)` |
+| ghost store dir | ✅ `type: ghost` | ⚠️ `--rpm-ghost`? (no deb equiv) | ❌ | ✅ `%ghost` |
+| systemd scriptlets | ✅ `scripts:` + macros? (plain scripts) | ✅ `--deb-systemd`, `--rpm-posttrans` | ✅ (deb) | ✅✅ native macros |
+| Prereqs on build host | Go binary (or brew/apt tarball) | Ruby gem | debhelper stack | rpm-build + mock |
+| Maintenance (2026) | 🟢 active (v2.47.0) | 🟢 active (v1.18.0) | 🔴 dormant since 2020 (Debian carries it) | 🟢 tool yes / 🔴 your spec |
+| Risk | young-ish config schema churn | docs thin | dead upstream | duplication forever |
+
+## Proposed pipeline (sketch)
+
+```make
+# Makefile additions (build only — install target stays for source installs)
+stage: dist/fenris-*.whl
+	rm -rf build/stage
+	python3 -m venv --copies build/stage/opt/fenris
+	build/stage/opt/fenris/bin/pip install --no-compile dist/fenris-*.whl
+	install -D -m 0755 scripts/fenris build/stage/usr/bin/fenris
+	install -D -m 0755 src/fenris/monitor.py  build/stage/usr/libexec/fenris/fenris-monitor
+	install -D -m 0755 src/fenris/collect.py  build/stage/usr/libexec/fenris/fenris-collect
+	install -D -m 0644 units/fenris-collect.timer   build/stage/usr/lib/systemd/system/fenris-collect.timer
+	install -D -m 0644 units/fenris-collect.service build/stage/usr/lib/systemd/system/fenris-collect.service
+	install -D -m 0644 polkit/com.bongbetic.fenris.monitor.policy \
+		build/stage/usr/share/polkit-1/actions/com.bongbetic.fenris.monitor.policy
+
+package-deb package-rpm: stage
+	nfpm pkg -f packaging/nfpm.yaml -p deb -t dist/
+	nfpm pkg -f packaging/nfpm.yaml -p rpm -t dist/
+```
+
+```yaml
+# packaging/nfpm.yaml (excerpt)
+name: fenris
+arch: amd64
+platform: linux
+version: ${VERSION}          # env expansion, documented feature
+maintainer: Fenris Maintainers <ops@bongbetic.com>
+description: SMART drive observation daemon with persistent TUI
+homepage: https://git.bongbetic.com/xavierk/Fenris
+depends: [smartmontools]
+contents:
+  - src: build/stage/        # everything above
+    dst: /
+    type: tree
+  - dst: /etc/fenris         # config dir; ship fenris.conf as config|noreplace
+    type: dir
+  - src: packaging/fenris.conf
+    dst: /etc/fenris/fenris.conf
+    type: config|noreplace
+  - dst: /var/lib/fenris     # rpm: %ghost ownership; deb: create in postinst
+    type: ghost
+scripts:
+  postinstall: packaging/postinst.sh   # groupadd fenris; install -d -o root -g fenris -m 2750 /var/lib/fenris; systemctl daemon-reload (units shipped dormant)
+  preremove:  packaging/prerm.sh        # stop timer if running
+overrides:
+  deb:
+    depends: [python3 (>= 3.9), smartmontools]
+  rpm:
+    depends: [python3 >= 3.9, smartmontools]
+```
+
+## Recommendation
+
+Adopt **nfpm + staged `--copies` venv**: closest to single source of truth
+(one YAML for both formats), smallest prerequisite surface (one static binary),
+actively maintained, and every Fenris constraint (units, polkit, libexec,
+`/etc/fenris` conffile, `/var/lib/fenris` ghost/store) has a first-class
+mapping. Keep fpm as documented fallback (identical staging tree, works
+anywhere Ruby exists). Do not build the release pipeline on dh-virtualenv
+(dormant, deb-only) or on a hand-maintained spec file (duplication, deb side
+unaddressed).
+
+## Sources
+
+- nfpm config reference: https://nfpm.goreleaser.com/configuration/ (source:
+  goreleaser/nfpm `www/content/docs/configuration.md`, accessed 2026-09-03)
+- nfpm CLI (single `-p`): goreleaser/nfpm `internal/cmd/package.go`
+- nfpm releases/status: GitHub API, repo pushed 2026-08-31, v2.47.0 2026-06-20
+- fpm README + CLI reference: https://github.com/jordansissel/fpm,
+  https://fpm.readthedocs.io/en/latest/cli-reference.html; releases v1.18.0
+  (2026-08-26), gem 1.18.0
+- dh-virtualenv docs: `doc/usage.rst`, `doc/tutorial.rst` @ master; tag 1.2.2
+  dated 2020-10-22 (GitHub commits API); PyPI project 404;
+  Debian 12 package 1.2.2-1.3 (packages.debian.org)
+- Distro textual versions: packages.debian.org, packages.ubuntu.com,
+  src.fedoraproject.org `python-textual.spec` f40–f43
+- Upgrade semantics: Debian Policy ch-files
+  (https://www.debian.org/doc/debian-policy/ch-files.html); Fedora packaging
+  guidelines (https://docs.fedoraproject.org/en-US/packaging-guidelines/);
+  RPM directive behavior quoted in nfpm config docs (%ghost, %config(noreplace))

docs/research/gitea-package-registry.md

@@ -0,0 +1,116 @@
+# Research: Gitea 1.27 Debian + RPM package registry feasibility
+
+Issue: [Fenris deb + rpm release plan](https://git.bongbetic.com/xavierk/Fenris/issues/33) →
+[Research: Gitea 1.27 Debian + RPM package registry feasibility](https://git.bongbetic.com/xavierk/Fenris/issues/35)
+Verified 2026-09-03 against live instance `https://git.bongbetic.com` (reports `1.27.1` via `/api/v1/version`)
+and primary sources: docs.gitea.com 1.27 Debian/RPM registry pages and Gitea `v1.27.1` source (go-gitea/gitea tag).
+
+**Verdict: feasible.** Every publish/consume path tested live with throwaway packages `fenris-regtest` (all deleted afterward; package list verified empty).
+
+## 1. Publish paths (verified live, HTTP 201)
+
+### Debian (`.deb`)
+
+```bash
+curl --user xavierk:$TOKEN --upload-file fenris_0.3.0_amd64.deb \
+  "https://git.bongbetic.com/api/packages/xavierk/debian/pool/{distribution}/{component}/upload"
+```
+
+- `distribution` and `component` are free-form path segments chosen at upload time (e.g. `bookworm/main`, `noble/main`). Gitea derives apt suites from what was uploaded — verified: same .deb published to `pool/bookworm/main` and `pool/noble/main` (both 201), both then served in `dists/bookworm/` and `dists/noble/` with correct `Suite:`/`Codename:` headers.
+- Republish of identical name+version+distribution+component+architecture → **409 Conflict** (verified). Must delete first.
+
+### RPM (`.rpm`)
+
+```bash
+# no group (flat repo)
+curl --user xavierk:$TOKEN --upload-file fenris-0.3.0-1.el9.x86_64.rpm \
+  "https://git.bongbetic.com/api/packages/xavierk/rpm/upload"
+# with group (distro tag, nestable)
+curl --user xavierk:$TOKEN --upload-file fenris-0.3.0-1.fc40.x86_64.rpm \
+  "https://git.bongbetic.com/api/packages/xavierk/rpm/el9/upload"   # e.g. el9, rocky/el9, fc40
+```
+
+- Group = free-form nesting used to partition repos per distro/track. Verified: publish to root group and `el9` group (both 201), duplicate → 409.
+- Owner can be the user (`xavierk`) or an org; packages under a public owner are readable anonymously (verified: metadata fetches without auth succeeded).
+
+## 2. Consumer setup (exact commands)
+
+### apt clients
+
+```bash
+sudo mkdir -p /etc/apt/keyrings
+sudo curl -o /etc/apt/keyrings/gitea-xavierk.asc \
+  https://git.bongbetic.com/api/packages/xavierk/debian/repository.key
+echo "deb [signed-by=/etc/apt/keyrings/gitea-xavierk.asc] https://git.bongbetic.com/api/packages/xavierk/debian bookworm main" \
+  | sudo tee /etc/apt/sources.list.d/gitea.list   # one line per distribution
+sudo apt update
+apt install fenris            # or fenris=0.3.0
+# private owner variant: https://{user}:{token}@git.bongbetic.com/api/packages/... in the URL
+```
+
+### dnf clients
+
+```bash
+sudo dnf config-manager --add-repo https://git.bongbetic.com/api/packages/xavierk/rpm/el9.repo
+# private owner: add user:token into the baseurl inside /etc/yum.repos.d/gitea-xavierk-el9.repo afterwards
+sudo dnf install fenris       # or fenris-0.3.0
+```
+
+The served `.repo` (verified live) sets `gpgcheck=1` and points `gpgkey` at `…/rpm/repository.key`, so `dnf` auto-imports on first use.
+
+## 3. Metadata signing: native, not passthrough
+
+Gitea **signs generated metadata itself** with per-instance auto-generated PGP keys. Client-side signing config is limited to trusting the served keys.
+
+- Debian: `dists/{suite}/InRelease` is clearsigned; `Release.gpg` detached sig also served. Key (RSA) fetched from `…/debian/repository.key`, uid literally `(Automatically generated Debian Registry Key; created …)`.
+- RPM: `repodata/repomd.xml.asc` detached ASCII-armored signature, uid `(RPM Registry)`. Key from `…/rpm/repository.key`.
+- Both verified with `gpg --verify` → **Good signature** (keys are self-generated; the "not certified" warning is expected and handled by the signed-by/keyring flow above).
+- The apt `Release` also advertises `Acquire-By-Hash: yes` with MD5/SHA1/SHA256/SHA512 indexes of `Packages`/`.gz`/`.xz` (verified live). RPM repomd carries sha256 checksums for `primary/filelists/other.xml.gz`.
+
+There is **no bring-your-own-signing-key config** for these registries in 1.27 — trust anchor is the instance's auto keys. For Fenris this is acceptable; TOFU over TLS via the key URLs above.
+
+## 4. Multi-distro metadata
+
+- Debian: distributions/suites are implicit — whatever `{distribution}` path segments appear on upload become `dists/{distribution}/` trees with `Suite:`/`Codename:` set to the segment. No server-side list to maintain; adding a new distro = upload with new segment + one more `deb …` sources line. Components likewise (`main`, etc.). Architectures come from each `.deb`'s control stanza (index served as `dists/{dist}/{component}/binary-{arch}/Packages`).
+- RPM: same via `{group}` path segments (`el9`, `rocky/el9`, …); each group gets its own `repodata/`. No `basearch` filtering — clients pick the group; Gitea publishes whatever RPM arch was uploaded.
+
+## 5. Version retention
+
+- Default: **all versions retained indefinitely**; nothing auto-deletes. Old versions stay installable (`apt install fenris=0.2.9`, `dnf install fenris-0.2.9`).
+- Republishing an existing name+version (deb: same dist/component/arch; rpm: same file name in group) → 409; overwrite requires delete-then-upload.
+- Optional cleanup rules exist (per owner + package type): `KeepCount`, `KeepPattern`, `RemoveDays`, `RemovePattern`, `MatchFullName` (source: `models/packages/package_cleanup_rule.go`, executed by scheduled `CleanupTask` in `services/packages/cleanup/cleanup.go`). In 1.27.1 they are configurable **only in the web UI** (owner → Packages → Cleanup Rules); no v1 REST route (verified by route table grep of `routers/api/v1/api.go` — probes of `/api/v1/packages/{owner}/cleanuprules…` return 404/409-style errors).
+- Deletes: format-specific `DELETE …/debian/pool/{dist}/{component}/{name}/{version}/{arch}` and `DELETE …/rpm/{group}/package/{name}/{version}/{arch}` (both verified, 204). Deleting last file removes the version. Generic fallback: `DELETE /api/v1/packages/{owner}/{type}/{name}/{version}`.
+
+## 6. Release attachment: not supported
+
+Gitea 1.27.1 has **no package↔release linkage**. Release assets (`…/releases/{id}/assets`) are standalone file uploads; the package model has no release field and no route links them (verified against `v1.27.1` source: `routers/api/v1/repo/release_attachment.go`, `models/packages/`). Options for Fenris releases:
+
+1. Publish `.deb`/`.rpm` to the registry (real apt/dnf install UX) and reference the registry URLs in release notes.
+2. Additionally upload tarballs/SHA256SUMS as plain release attachments.
+3. Generic registry (`PUT /api/packages/{owner}/generic/{name}/{version}/{filename}`) if an untyped artifact store is needed.
+
+## 7. Caveats for the release plan
+
+- Owner choice matters: publish under an **org** (e.g. `fenris`) if multiple maintainers need write; `xavierk` user owner works today (token owner is admin).
+- Metadata access follows owner visibility — public owner → anonymous consumers, no token in URLs (current state, verified). Keep owner public for frictionless installs, or embed `user:token` in sources/baseurl.
+- apt distro naming should match OS release names (`bookworm`, `trixie`, `noble`) purely for client convention; server accepts anything.
+- RPM groups should mirror `$distver` (e.g. `el9`, `fc40`) so `.repo` selection is obvious per target.
+
+## 8. Test log (live, 2026-09-03)
+
+| Step | Result |
+|---|---|
+| `PUT debian/pool/bookworm/main/upload` | 201 |
+| `PUT debian/pool/noble/main/upload` (multi-dist) | 201 |
+| `PUT debian` duplicate | 409 (expected) |
+| `PUT rpm/upload` (no group) | 201 |
+| `PUT rpm/el9/upload` (group) | 201 |
+| `PUT rpm` duplicate | 409 (expected) |
+| `GET debian/repository.key` / `rpm/repository.key` | PGP public keys (200) |
+| `GET dists/bookworm/{Release,InRelease,Packages}` | correct; `gpg --verify` Good signature |
+| `GET rpm{,/el9}/repodata/repomd.xml{,.asc}` | 200; Good signature |
+| `GET rpm{,/el9}.repo` | generated repo files with `gpgcheck=1` |
+| Cleanup-rules REST probes | 404 (not in v1 API — UI only) |
+| `DELETE` all four test entries | 204 ×4; package list then empty |
+
+Sources: [docs.gitea.com 1.27 Debian registry](https://docs.gitea.com/1.27/usage/packages/debian), [docs.gitea.com 1.27 RPM registry](https://docs.gitea.com/1.27/usage/packages/rpm), Gitea source tag `v1.27.1` (`routers/api/v1/api.go`, `models/packages/package_cleanup_rule.go`, `services/packages/cleanup/cleanup.go`), live instance `git.bongbetic.com`.

docs/research/obs-route.md

@@ -0,0 +1,73 @@
+# Research: OBS as an alternative build + distribution route
+
+Resolves [Research: OBS as alternative build + distribution route](https://git.bongbetic.com/xavierk/Fenris/issues/36) on the [Wayfinder map](https://git.bongbetic.com/xavierk/Fenris/issues/33).
+
+- Date: 2026-09-03
+- Verdict: **Reject OBS now; ship via the self-hosted Gitea 1.27.1 registry** (deb + rpm), and revisit OBS only if publishing reach becomes a goal.
+
+## Question
+
+Evaluate openSUSE Open Build Service (OBS) as the build + distribution route — deb build quality, vendoring `textual>=0.40` via source services (offline sandbox), signing, publishing reach, account/maintenance cost, build latency — against the Gitea registry on our matrix: Debian 12, Ubuntu 22.04/24.04, Fedora 40+, x86_64.
+
+## Findings
+
+### 1. deb build support quality — real, with quirks
+
+- OBS builds deb via the classic recipe trio: `debian.control`, `debian.rules`, `PACKAGE.dsc` (OBS User Guide §2.3 "Debian: Dsc"). The build phase runs `dpkg-buildpackage` on Debian-based distributions (§25.1.3 "Package Build"); Debian build environments can alternatively use the `debootstrap` build engine (§"Configuration File Syntax", `BuildEngine`).
+- Quirk: release numbers are **not** auto-incremented across rebuilds unless the dsc carries `DEBTRANSFORM-RELEASE` (§2.3) — a packaging decision we'd own either way.
+- Upstream build deps are available: `dh-virtualenv` and `dh-python` exist in Debian 12 (packages.debian.org, checked 2026-09-03), so the ADR-0004 venv/lockfile design maps onto an OBS dsc without patching the build root.
+- All five matrix targets exist as public OBS build roots: `Debian:12`, `Ubuntu:22.04`, `Ubuntu:24.04`, `Fedora:40`, `Fedora:41` — each project `_meta` answered HTTP 200 on build.opensuse.org (checked 2026-09-03).
+- Live proof of deb publishing quality: `isv:ownCloud:desktop/Debian_10` on download.opensuse.org serves a proper Debian archive (`Release`, `Release.gpg`, `InRelease` all HTTP 200, checked 2026-09-03).
+
+### 2. Vendoring textual≥0.40 — the offline sandbox forces the same work we already planned
+
+- The build environment has **no network**: "services requiring external network access are likely to fail in [buildtime] mode, because such access is not available if the build workers are running in secure mode (as is always the case at https://build.opensuse.org)" (User Guide §7.2, "Modes of Source Services"); Dockerfile builds likewise run "in a safe build environment without network access" (§29.3).
+- Vendoring must therefore happen **before** the build, via source services that run server-side on commit (`default`/`trylocal` modes, §7.2) or via files committed to the package. The standard services are per-file fetchers — `download_url` (§22.1.3), `download_files`, `obs_scm`/`tar`/`set_version` (§8 SCM integration) — there is no "pip resolve" service, so a pinned dependency tree like Fenris's means either N `download_url` entries mirroring the committed lockfile, or simply committing the vendored wheel/sdist tree.
+- Conclusion: OBS does not remove the vendoring step; it reproduces ADR-0004's committed-lockfile design with extra XML. Since distro `python3-textual` is 0.1.13 on Debian 12, Ubuntu 22.04 and 24.04 (packages.debian.org / packages.ubuntu.com, checked 2026-09-03) — far below the `>=0.40` floor — vendoring is unavoidable on any route.
+
+### 3. Signing — OBS key, not ours; Gitea deb repo is our key
+
+- OBS signs published repositories with the **instance's** key: one signer per partition "calls an external tool to execute the signing" (User Guide §23 "OBS Architecture", Signer); consumers accept the OBS repo key ("When prompted, accept the GPG key of the download repository", §1.10). A build.opensuse.org user cannot upload a personal signing key. Trust therefore flows to openSUSE infra, and the signature says nothing about Fenris's maintainers.
+- Gitea 1.27.1's Debian registry serves apt metadata signed with the Gitea instance's PGP key (`repository.key` endpoint, `signed-by` in sources.list — docs.gitea.com, "Debian Package Registry"), i.e. **our** host and **our** key. The RPM registry serves a `.repo` endpoint but documents no GPG signing of repodata; rpm-file signing stays our choice at build time.
+
+### 4. Publishing reach — OBS wins reach; reach is not our bottleneck
+
+- OBS publishes home-project results to `https://download.opensuse.org/repositories/home:USER/<dist>` (§1.10) and offers generated download pages on software.opensuse.org (§17.4). That is genuine CDN-class reach.
+- Caveats from the same docs: branched projects are **not** published by default (§1.10), and the repo is a live view of the project state — no release artefact pinning; deleting the project or flag disables distribution.
+- The Gitea route's reach is exactly `git.bongbetic.com` plus whatever the README says — adequate for a named four-distro matrix whose users follow our instructions, and it keeps the release artefact under versioned control on the same host as the source.
+
+### 5. Account and maintenance cost — strictly additive
+
+- Using build.opensuse.org requires an openSUSE account (single sign-on; the web UI's "Sign up!") and work happens in `home:USERNAME` plus permitted subprojects (§"Setting Up Your Home Project for the First Time"; §23 "OBS Concepts" on home projects).
+- Day-to-day: `osc` + `_service` XML + dsc/spec recipes maintained in OBS's own package VCS, kept in sync with Fenris's git. The SCM bridge (`scmsync`) does support self-hosted Gitea ("We also support Self-Hosted instances from GitHub, GitLab and Gitea", §8.1.3; setup in §28.1.2 — build descriptions must live in the repo's top level), but it also disables OBS-side workflows (no `_link` merging, limited workflows, §28.1.1).
+- No published quota/SLA for the public instance; capacity and availability are a shared commons. The Gitea route needs zero new accounts, zero new artefact formats beyond the two package recipes we must write anyway, and reuses the existing release host.
+
+### 6. Build latency — shared queue vs. deterministic local
+
+- OBS routes every commit through scheduler → dispatcher → shared workers; the dispatcher "tries to assign jobs fairly between the project repositories" using a per-repository load model (§23, Scheduler/Dispatcher). For our five tiny x86_64 jobs this is typically minutes, but there is no documented SLA and the queue is global — worst case is unbounded (estimate; the docs guarantee only fairness, not latency).
+- The Gitea route builds wherever `make` runs and publishes with one authenticated `PUT` per artefact (docs.gitea.com: Debian `PUT .../pool/{distribution}/{component}/upload`; RPM `PUT .../rpm/{group}/upload`). Latency = build time, fully under our control.
+
+## Comparison on the 4-distro matrix
+
+| Axis | OBS (build.opensuse.org) | Gitea 1.27.1 registry |
+|---|---|---|
+| Debian 12 / Ubuntu 22.04/24.04 deb | dsc + dpkg-buildpackage; DEBTRANSFORM-RELEASE quirk | we build the same deb locally, upload via PUT |
+| Fedora 40+ rpm | spec + rpmbuild in Fedora roots | same spec built locally, `.repo` grouping (`fedora/40`) |
+| Vendoring textual≥0.40 | offline sandbox forces committed vendored tree (no pip service) | same committed vendored tree (ADR-0004 lockfile) |
+| Signing | OBS instance key (not ours) | deb repo signed with our key; rpm repodata unsigned |
+| Reach | download.opensuse.org CDN + software.o.o pages | our domain only |
+| Accounts/infra | new openSUSE account, osc workflow, commons SLA-free | zero new infra |
+| Latency | global shared queue, minutes typical, no SLA | deterministic (local build) |
+
+## Recommendation
+
+**Reject OBS as the build + distribution route for Fenris.** The offline sandbox forces the exact vendoring work the Gitea route already requires, so OBS adds cost (account, osc/source-service maintenance, external commons in the release path, queue latency) without removing any; its one real advantage — CDN and software.o.o reach — does not matter for a hobby project whose four target distros are served by one signed apt repo and one rpm repo on the existing Gitea host, under our own key.
+
+Revisit trigger: if Fenris later wants one-click installs via software.opensuse.org, architectures beyond x86_64, or many more distro targets — the deb publishing quality (verified live) and self-hosted-Gitea SCM bridge make OBS a viable amplifier then.
+
+## Sources
+
+- OBS User Guide (openbuildservice.org/help/manuals/obs-user-guide/, PDF): §2.3 Debian: Dsc; §7 Using Source Services (offline buildtime services, modes); §8.1.3 Supported SCMs; §17.4 download pages; §22.1.3 download_url; §23 OBS Architecture (Scheduler/Dispatcher/Signer); §25.1.3 Package Build; §28.1 SCM bridge; §29.3 Dockerfile builds (no network); §1.10 Installing Packages from OBS; "Configuration File Syntax" (BuildEngine, Repotype: debian).
+- Live checks (2026-09-03): `Debian:12`/`Ubuntu:22.04`/`Ubuntu:24.04`/`Fedora:40`/`Fedora:41` project `_meta` on build.opensuse.org (all 200); `isv:ownCloud:desktop/Debian_10` `Release`/`Release.gpg`/`InRelease` on download.opensuse.org (all 200).
+- packages.debian.org / packages.ubuntu.com (2026-09-03): `python3-textual` 0.1.13 on bookworm, jammy, noble; `dh-virtualenv`, `dh-python` present in bookworm.
+- docs.gitea.com, "Debian Package Registry" and "RPM Package Registry" (1.27 line): apt sources with `signed-by` + `repository.key`, `PUT` upload endpoints, `.repo` groups.

docs/spec/acceptance-criteria.md

@@ -91,7 +91,7 @@ Status: Accepted — resolves [Define cross-cutting acceptance criteria](https:/
 - **TUI-1** (A) Variant A "Panes": one dense keyboard-first screen; confidence rendered as evidence (state + contributing facts); boot enablement, runtime activity, last collect outcome, and freshness displayed as four separate facts.
 - **TUI-2** (M) Pause/resume asymmetry and polkit tty passthrough work in a live terminal: pause confirms, resume does not, and the platform agent prompts without breaking the TUI.
 - **TUI-3** (P) Textual runs on Python 3.9+, gated at install time, never a runtime crash.
-- **TUI-4** (A) The Panes screen layout is normative: a full-width headline band (lifespan headline or its no-projection wording, confidence state with contributing facts, scenario range); a usage-history pane on the left (write-history sparkline with ▲ habit-change and ? unexplained-gap markers plus legend, habit-split bar with active/idle/powered-off/unknown shares); a drive-health and settings pane on the right (health facts, vendor-wear context line, read-only settings with the endurance baseline and its provenance label); a full-width service strip at the bottom (the four separate service facts, the monitoring-period line, the action legend). Production bindings are `p` pause (asks), `r` resume (does not), `c` collect now, `d` disclosures, `q` quit ([Prototype the TUI information architecture](https://git.bongbetic.com/xavierk/Fenris/issues/3)); the prototype branch is visual reference only.
+- **TUI-4** (A) The Panes screen layout is normative: a full-width headline band (lifespan headline or its no-projection wording, confidence state with contributing facts, scenario range); a usage-history pane on the left (write-history sparkline with ▲ habit-change and ? unexplained-gap markers plus legend, habit-split bar with active/idle/powered-off/unknown shares); a drive-health and settings pane on the right (health facts, vendor-wear context line, read-only settings with the endurance baseline and its provenance label); a full-width service strip at the bottom (the four separate service facts, the monitoring-period line, the action legend). Production bindings are the footer `p pause · r resume · c collect · d disclosures` — pause asks, resume does not — plus a bordered quit rail `q QUIT TUI` visually separate from monitoring state; the rail owns quit and the footer carries no quit entry (bindings amended by [Lock the dashboard wording strings](https://git.bongbetic.com/xavierk/Fenris/issues/57); original [Prototype the TUI information architecture](https://git.bongbetic.com/xavierk/Fenris/issues/3)); the prototype branch is visual reference only.
 
 ## Failure and recovery (ADR 0005)
 
@@ -117,6 +117,13 @@ Status: Accepted — resolves [Define cross-cutting acceptance criteria](https:/
 - **IN-9** (P) The installer verifies `python3 ≥ 3.9` and fails cleanly otherwise; `/var/lib/fenris` is created with root-written group-read permissions; the database file is created lazily by the first write.
 - **IN-10** (P) Installed artifacts sit only at their fixed locations — units in `/etc/systemd/system`, helpers in `/usr/libexec/fenris`, polkit policy under `/usr/share/polkit-1/actions/`, configuration at `/etc/fenris`, observation store under `/var/lib/fenris` — and every placed file is recorded in the manifest (ADR 0004 §2; ADR 0003 §4).
 
+## Migration from make-install systems (ADR 0007 §10, spec §9)
+
+- **MG-1** (M) The migration runbook is published in the install docs (`docs/install/migrate-from-makeinstall.md`): mandatory remove-then-install steps, why over-install is forbidden (stale admin-directory units silently shadow vendor units; the local wrapper shadows the package wrapper), no-move continuity, and the reset-to-dormant expectation (the user opts back in with the sanctioned resume).
+- **MG-2** (A) The install guard is verified across the matrix: either make-install marker (the legacy placement manifest, or a unit file under the admin unit directory) causes an abort with a runbook pointer — never auto-clean. Tested by `test_migration_guard` on all four targets (Debian 12, Ubuntu 22.04, Ubuntu 24.04, Fedora 40).
+- **MG-3** (A) No-move continuity is verified in a container seeded with a make-install-shaped system: existing group makes sysusers a no-op, existing store directory makes tmpfiles a no-op, the hand-written configuration survives as a non-database file (package default lands beside it), and the store schema is caught up by the upgrade-path migration. Tested by `test_no_move_continuity_deb` and `test_no_move_continuity_rpm`.
+- **MG-4** (M) The migration costs at most one short sample gap, honestly recorded in the endurance timeline: `make uninstall`'s sanctioned disable closes the open period `user_disabled`; after migration the user opts back in with `fenris monitor resume`.
+
 ## Collector acquisition path (ADR 0006)
 
 - **AC-1** (P) Each collection run acquires counters and thermal evidence solely from `smartctl -a -j <device>` and controller identity (`subnqn`, `sn`, `mn`, `fr`, `transport`) solely from sysfs; no other acquisition path exists anywhere in the codebase.
@@ -124,3 +131,16 @@ Status: Accepted — resolves [Define cross-cutting acceptance criteria](https:/
 - **AC-3** (A) Any acquisition failure — missing binary, nonzero exit, malformed JSON, unreadable sysfs attribute — fails the whole collection run; no partial sample (identity without counters, or counters without identity) is ever written; the miss surfaces through ADR 0005 freshness, never as degraded identity.
 - **AC-4** (P) `vid`/`ssvid` are read from the PCI sysfs node when present and stored null otherwise; they are segment metadata only, never key components.
 - **AC-5** (P) `make install` verifies `smartctl` and fails cleanly otherwise; the acquisition path adds no Python dependency and no OS package beyond smartmontools (ADR 0004 §9).
+
+## Dashboard clarity and release notes ([Chart Fenris dashboard clarity](https://git.bongbetic.com/xavierk/Fenris/issues/55))
+
+Decided in [Write the dashboard clarity acceptance criteria](https://git.bongbetic.com/xavierk/Fenris/issues/59), from [Prototype the dashboard clarity additions](https://git.bongbetic.com/xavierk/Fenris/issues/56), [Lock the dashboard wording strings](https://git.bongbetic.com/xavierk/Fenris/issues/57), and [Specify the changelog and release-notes mechanism](https://git.bongbetic.com/xavierk/Fenris/issues/58).
+
+- **DC-1** (A) TUI branding: the header bar renders `Fenris — NVMe endurance monitor`; a dimmed `by Bongbetic` sits inline with service facts in the bottom service strip; neither string appears in `fenris status` (TUI-only identity surfaces).
+- **DC-2** (A) Continuity parity, keyed to the boot fact as-is: active + boot-enabled renders `monitoring: active in background · persists across reboots`; boot-disabled renders `monitoring: does not start on next boot` — identical lowercase source strings in the TUI service strip and `fenris status`, including while paused (paused implies boot-disabled; the row still reports the fact). Test impact: feeds the CI-2 sweep (lowercase source-string comparison).
+- **DC-3** (A) Paused presentation (Deliberate disable): the TUI shows a strong state block titled `monitoring: paused — deliberate disable` with subline `paused time is excluded from your usage habit · resume: fenris monitor resume`; `fenris status` prints the same two lines with identical wording. Test impact: feeds the CI-2 sweep (lowercase source-string comparison).
+- **DC-4** (A) Quit affordance distinct from monitoring state: a bordered labelled rail `q QUIT TUI` visually separate from the paused state block; the footer reads `p pause · r resume · c collect · d disclosures` with no quit entry (the rail owns quit); quitting the TUI never alters monitoring state. Amends TUI-4's binding parenthetical.
+- **DC-5** (A) Launch auth banner: `privileged actions will prompt for authentication (polkit)` renders full-width under the header at TUI launch, clears on the first refresh tick, and never reappears in the session; no user-facing string uses "sudo" (polkit-accurate elevation wording only).
+- **DC-6** (A) CHANGELOG.md shape (Keep a Changelog 1.1): `## [Unreleased]` always present at top, even empty; version headings `## [X.Y.Z] - YYYY-MM-DD` with strict ISO date; categories Added/Changed/Fixed only, security folding into Fixed; entries are single `- ` bullets, imperative mood, user-facing, no commit hashes or issue numbers.
+- **DC-7** (A) Extraction fails closed: `scripts/extract_changelog.py` slices the requested version's section verbatim and never reads `[Unreleased]`; a missing or empty section or a malformed date produces `::error::` and a nonzero exit; the release workflow fails when the pushed tag ≠ `v{version from pyproject.toml}` (guard skipped on `workflow_dispatch`).
+- **DC-8** (A/P) Release body: the body is the extracted section verbatim plus the standing footer from `packaging/release-footer.md`; a re-run against an existing release PATCHes the body (re-sync is a feature) while uploaded assets skip idempotently. A covers assembly/PATCH-logic unit tests; P is one scripted `workflow_dispatch` verification of body assembly.

docs/spec/dashboard-clarity.md

@@ -0,0 +1,122 @@
+# Fenris dashboard clarity specification
+
+**Status: decision-complete.** Assembled by [Assemble the dashboard clarity specification and close the map](https://git.bongbetic.com/xavierk/Fenris/issues/60) from the closed tickets of the Wayfinder map [Chart Fenris dashboard clarity](https://git.bongbetic.com/xavierk/Fenris/issues/55). This document is normative for the follow-up **execution effort**; nothing here is implemented by the map.
+
+**Canonical roles.** The [redesign specification](fenris-redesign.md) (frozen) and [ADRs 0001–0007](../adr/) remain authoritative and untouched — this is a companion spec covering five dashboard clarity additions plus the changelog-driven release-notes mechanism. The [criteria register](acceptance-criteria.md) carries the testable statements: **DC-1–DC-8**, appended by this assembly, with **TUI-4's binding list amended** (§4). Terminology follows the glossary in [`CONTEXT.md`](../../CONTEXT.md), including *Deliberate disable* and *Release*.
+
+**Binding language.** *Must*, *exactly*, and *never* are normative.
+
+## How to read this document
+
+Five screen additions (§1–§5), one release-notes mechanism (§6), the verbatim string register (§7), and the README section to add at execution (§8). Each section cites its criteria. Source strings are lowercase; the TUI may render uppercase via styling only. Typography, governing every string: em-dash `—` separates a title from its qualifier; middle dot `·` joins facts within a line; UTF-8 is assumed. CI parity sweeps compare lowercase source strings — rendering case is styling, not wording.
+
+## 1. Header bar and Bongbetic credit — DC-1
+
+Visual base is treatment A, quiet integration: the existing Panes information architecture is preserved.
+
+- The header bar reads `Fenris — NVMe endurance monitor`.
+- The credit `by Bongbetic` renders dimmed, inline with service facts in the bottom service strip — never in the action row.
+- Both are TUI-only identity surfaces: `fenris status` never renders them.
+
+## 2. Continuity line — DC-2
+
+A labelled `CONTINUITY` row in the service strip (treatment B), mirrored by `fenris status` — the TUI/CLI parity anchor. The row is keyed to the boot fact as-is, independently of run state (Deliberate disable runs `systemctl disable --now`, so paused implies boot-disabled; the row still reports the fact):
+
+- Active + boot enabled: `monitoring: active in background · persists across reboots`
+- Boot disabled: `monitoring: does not start on next boot`
+
+Identical lowercase source strings in the TUI service strip and `fenris status`, including while paused.
+
+## 3. Paused state block — DC-3
+
+Treatment C, strong state blocks: when monitoring is paused, a full-width, high-contrast banner clearly identifying Deliberate disable:
+
+- Title: `monitoring: paused — deliberate disable`
+- Subline: `paused time is excluded from your usage habit · resume: fenris monitor resume`
+
+`fenris status` prints the same two lines with identical wording (state line + consequence line). The resume hint uses the CLI form only; the footer owns key hints — no duplication.
+
+## 4. Quit rail — DC-4 (amends TUI-4)
+
+Treatment B, labelled rails: a prominent bordered `q QUIT TUI` rail, visually separate from the monitoring-state block and the paused banner. The footer becomes `p pause · r resume · c collect · d disclosures` — the rail owns quit; the footer carries no quit entry. Quitting the TUI never alters monitoring state. The register's TUI-4 binding parenthetical is amended accordingly by this assembly.
+
+## 5. Launch auth banner — DC-5
+
+A quiet informational line (treatment A) that never competes with drive state:
+
+- Text: `privileged actions will prompt for authentication (polkit)`
+- Full-width under the header at TUI launch; clears on the first refresh tick; never reappears in the session.
+- TUI-only; `fenris status` never shows it.
+- Elevation wording is polkit-accurate everywhere: no user-facing string uses "sudo" (sudo belongs to install/upgrade docs).
+- Evidence class A: a Textual pilot drives refresh ticks headlessly.
+
+## 6. Changelog and release notes — DC-6, DC-7, DC-8
+
+Implements the existing glossary term *Release* (tag + packages + change notes together). No new glossary terms; no ADR (reversible mechanism).
+
+### 6.1 CHANGELOG.md (source of truth, repo root)
+
+- Keep a Changelog 1.1 shape. `## [Unreleased]` is always present at top, even empty. Version headings are `## [X.Y.Z] - YYYY-MM-DD` — bracketed bare semver, strict ISO date.
+- Categories are `### Added`, `### Changed`, `### Fixed` only; security fixes fold into Fixed.
+- Entries are single `- ` bullets, imperative mood, user-facing phrasing; no commit hashes or issue numbers.
+
+### 6.2 Extraction (release.yml, tag time)
+
+- `scripts/extract_changelog.py` (checked in, unit-tested): takes the changelog path and a version; slices that version's section verbatim; never reads `[Unreleased]`. Fails closed — `::error::` plus nonzero exit — when the section is missing or empty or the date is malformed.
+- Guard: the workflow fails when the pushed tag ≠ `v{version from pyproject.toml}` (guard skipped on `workflow_dispatch`).
+
+### 6.3 Release body
+
+- Body = extracted version section verbatim + standing footer from `packaging/release-footer.md` (channel install one-liners, `sha256sum -c SHA256SUMS.asc` verify, rollback pointer). The footer is standing text; only the changelog section varies.
+- Re-run against an existing release: PATCH the body (changelog re-sync is a feature); uploaded assets/packages keep their current idempotent-skip.
+
+### 6.4 Discipline
+
+- All entries land in `[Unreleased]` as part of the fixing change — no notes-later step.
+- One release commit bumps the pyproject version, renames `[Unreleased]` → the version heading, and restores an empty `[Unreleased]`; the tag points at that commit (tag ↔ pyproject ↔ changelog triple-match, enforced fail-closed by DC-7).
+- No backfill: per-release notes begin with the release shipping this mechanism; `CHANGELOG.md` starts with empty `[Unreleased]`.
+
+## 7. String register (verbatim)
+
+### TUI-only strings (launch/identity surfaces)
+
+| Surface | String |
+|---|---|
+| Header bar | `Fenris — NVMe endurance monitor` |
+| Credit (dimmed, inline with service facts) | `by Bongbetic` |
+| Auth banner (full-width under header at launch, clears on first refresh tick, never reappears) | `privileged actions will prompt for authentication (polkit)` |
+| Quit rail (bordered, labelled) | `q QUIT TUI` |
+| Footer (owns key hints; no quit entry) | `p pause · r resume · c collect · d disclosures` |
+
+### Parity strings (TUI and `fenris status` identical — CI-2)
+
+| Surface | String |
+|---|---|
+| Continuity, active + boot enabled | `monitoring: active in background · persists across reboots` |
+| Continuity, boot disabled | `monitoring: does not start on next boot` |
+| Paused state line | `monitoring: paused — deliberate disable` |
+| Paused consequence line | `paused time is excluded from your usage habit · resume: fenris monitor resume` |
+
+`fenris status` prints the paused state line + consequence line when paused, identical wording to the banner title + subline.
+
+## 8. README section (add at execution)
+
+The README gains a "Reading the dashboard" section after the CLI reference. Verbatim text:
+
+```markdown
+## Reading the dashboard
+
+`fenris` opens the TUI dashboard. Three things it tells you:
+
+- **Continuity** — the service strip's continuity line (and `fenris status`) reports whether monitoring survives reboots: `monitoring: active in background · persists across reboots`, or `monitoring: does not start on next boot`.
+- **Paused vs. quit** — a full-width `monitoring: paused — deliberate disable` block means collection is stopped (`fenris monitor pause`); resume with `fenris monitor resume`. Pressing `q` only leaves the screen — monitoring keeps running in the background.
+- **Auth banner** — at launch, `privileged actions will prompt for authentication (polkit)` shows once and clears on the first refresh. Privileged actions elevate via polkit; Fenris never asks for sudo.
+
+Per-release notes live on the [releases page](https://git.bongbetic.com/xavierk/Fenris/releases): each entry is the version's `CHANGELOG.md` section — what was added, changed, and fixed — plus standing install and verification instructions.
+```
+
+This resolves the map's README-wording fog: the wording is decided here; the actual README edit is execution.
+
+## 9. Out of scope
+
+Executing any of this — code, tests, releases — and any TUI layout or information-architecture redesign beyond the five additions named above. Execution is a fresh effort after handoff.

docs/spec/release-packaging.md

@@ -0,0 +1,106 @@
+# Fenris release and packaging specification
+
+**Status: decision-complete.** Assembled by [Task: Compose release spec + ADR amending 0004](https://git.bongbetic.com/xavierk/Fenris/issues/42) from the closed tickets of the Wayfinder map [Fenris deb + rpm release plan](https://git.bongbetic.com/xavierk/Fenris/issues/33). This document is normative for the follow-up **execution effort** that builds and publishes packages; no packages are built here.
+
+**Canonical roles.** [ADR 0007](../adr/0007-package-delivery-amends-0004.md) records the lifecycle rationale (amending [ADR 0004](../adr/0004-install-upgrade-removal-lifecycle.md)); this document restates the **operative contracts** — compat matrix, channel, toolchain, signing, release mechanics, package layout, maintainer-script behavior, migration — so the executing effort never needs Wayfinder-ticket access. Runtime semantics come from [ADRs 0001–0006](../adr/) and the [redesign specification](fenris-redesign.md) verbatim; nothing here overrides them. Terminology follows the glossary in [`CONTEXT.md`](../../CONTEXT.md), including *Release* and *Rollback*.
+
+**Binding language.** *Must*, *exactly*, and *never* are normative.
+
+## 1. Compatibility matrix
+
+| Target | Version | Format | Registry placement |
+|---|---|---|---|
+| Debian 12 (bookworm) | — | deb | `debian/pool/bookworm/main` |
+| Ubuntu 22.04 (jammy) | — | deb | `debian/pool/jammy/main` |
+| Ubuntu 24.04 (noble) | — | deb | `debian/pool/noble/main` |
+| Fedora 40+ | every release | rpm | `rpm/fenris` group |
+| openSUSE Tumbleweed | rolling | rpm | `rpm/fenris` group |
+
+- Architecture: **x86_64 only** (arm64 only if real ARM hardware appears — map fog).
+- Dependencies are vendored as locked, pure-Python runtime packages for every target: Debian 12 and Ubuntu 22.04/24.04 ship `python3-textual` 0.1.13, far below the floor; Fedora 40+ ships ≥ 0.48 but below the pin ([toolchain research](../research/deb-rpm-toolchain.md)). No distro `python3-textual` dependency ever enters package metadata.
+- Package metadata `depends:`/`Requires:` are exactly `python3 (>= 3.10)`, `smartmontools`, `systemd` — the Python floor is 3.10 (oldest supported distro interpreter, Ubuntu 22.04), bumping ADR 0004 §10's 3.9 gate for packages; `make install` keeps the checkout's floor.
+- Runtime packages are staged with `python3 -m pip --target /opt/fenris/vendor`; entry points run the target system's `python3` with that directory on the import path. No package ships a copied Python interpreter, avoiding build-host ABI paths and rolling-distribution minor-version breakage.
+
+## 2. Distribution channel
+
+- **Channel:** the self-hosted Gitea 1.27.1 package registry at `git.bongbetic.com`, owner public for anonymous consumers ([registry research](../research/gitea-package-registry.md); [OBS rejected](../research/obs-route.md)).
+- **Single channel.** No stable/testing split — deferred until external users ask to track pre-release builds (map fog). Every published version is retained indefinitely (registry has no REST cleanup; republishing a filename is a 409).
+- **deb publication:** one deb artifact PUT to each codename pool — `PUT /api/packages/{owner}/debian/pool/{bookworm|jammy|noble}/main/upload`.
+- **rpm publication:** one rpm artifact PUT to the single `fenris` group — `PUT /api/packages/{owner}/rpm/fenris/upload` — serving Fedora 40+ collectively.
+- **Consumer setup (install docs, normative):**
+  - apt: keyring file from `…/debian/repository.key` via `signed-by`, one sources line per distribution, instance Debian Registry Key **fingerprint printed beside the curl one-liner** (TOFU hardening).
+  - dnf: `dnf config-manager --add-repo <raw-url of packaging/fenris.repo>` — the in-repo, Fenris-owned `.repo` with `gpgkey` pointing at the published packaging key and `repo_gpgcheck=0`. **Gitea's auto-generated `.repo` is never mentioned in docs**: it sets `gpgcheck=1` against the instance auto-key, which never signed our rpm payload — a trap that breaks installs.
+
+## 3. Build toolchain
+
+- **nfpm** for both formats from a single `packaging/nfpm.yaml` — one config, `overrides:` for per-format deltas, two invocations (`nfpm pkg -p deb`, `nfpm pkg -p rpm`). fpm is dropped entirely (CLI-flag config drifts); no hand rpm spec; dh-virtualenv is deb-only and dormant since 2020.
+- **Single source of truth:** version injected from `pyproject.toml`; file lists generated by a staging script (locked runtime packages → `/opt/fenris/vendor`, plus wrapper, helpers, units, polkit policy, sysusers/tmpfiles fragments) referenced by `nfpm.yaml` as a `type: tree` content entry — no hand-maintained file lists.
+- **Entry point:** `make package` → `dist/fenris_<v>_amd64.deb` + `dist/fenris-<v>-1.x86_64.rpm`.
+- **Version scheme:** `<pyproject-version>-1` in both formats; a rebuild of the same upstream version bumps the revision (`-2`, `-3`, …) — the same filename is never re-PUT (registry 409s duplicates).
+- **Authoring `nfpm.yaml`, the staging script, and the workflow file is execution** — deliberately not part of the decision map. The [toolchain research doc](../research/deb-rpm-toolchain.md) sketches the pipeline.
+
+## 4. Signing and key policy
+
+- **RPM payload: signed.** rpmsign with the dedicated packaging key, invoked by `make sign-rpm` after the package is built. This is required, not optional: it is the only working dnf-native verification path.
+- **deb: unsigned.** apt never verifies payload signatures; trust = instance-signed `InRelease` (signed-by keyring) + TLS + Acquire-By-Hash. Manual-download integrity is covered by SHA256SUMS.
+- **SHA256SUMS: clearsigned** with the packaging key — the trust anchor for manually downloaded release assets, independent of TLS.
+- **Packaging key:** single dedicated key, RSA 3072, UID `Fenris Packaging <packaging@bongbetic.com>`, 2-year expiry, no master/subkey hierarchy (single maintainer, manual builds). Private key lives in the password manager only; each release does import → sign → delete — nothing permanent on any build host. The full ceremony is documented in `docs/install/signing-key-ceremony.md`.
+- **Public key publication:** in-repo `packaging/keys/fenris-packaging.asc` (raw URL doubles as the `.repo` gpgkey target), release notes, docs page. No keyservers — TOFU-over-TLS.
+- **Rotation (outline):** new key published alongside old; rpm signed with the new key; `fenris.repo` gpgkey lists both URLs (dnf accepts multiple); old key dropped after one release cycle. Procedure details stay in map fog.
+
+## 5. Release mechanics
+
+- **A Release is:** a version tag, its packages in the channel, a Gitea release entry with notes, and a clearsigned SHA256SUMS — all together. **Bare tags are forbidden** (tag without packages + release entry is not a Release).
+- **Cadence: on-demand.** Tag when user-visible changes or fixes accumulate; no calendar, no empty releases, no frequency SLA, no RC ceremony — fixes ship as a revision bump of the current version.
+- **Versioning: plain semver.** Major = breaking CLI/config/unit change; store schema changes ride the natural bump (the forward-only refusal handles old-reader/new-store).
+- **Promotion flow:** tag → `make release` (automated: `make package` → RPM signing via nfpm → SHA256SUMS generation → clearsign → prints registry PUTs + Gitea release steps). The ceremony is documented in `docs/install/signing-key-ceremony.md`.
+- **Rollback:** installing an older package over a newer store is **unsupported** — the store's forward-only version refusal fails it by design. Documented rollback = restore the observation-store snapshot, then install the old Release. No automatic downgrade machinery exists or will be built.
+- **CI:** no runners are registered on the instance today ([Actions runner research](https://git.bongbetic.com/xavierk/Fenris/issues/37)), so the manual flow above is primary. A dormant `.gitea/workflows/release.yml` (`on: push: tags: ['v*']`, single job, host-mode runner) is committed alongside; if it fires, it replicates `make release`. Cheapest future upgrade: one `act_runner` static binary in host-label mode on the existing Gitea host.
+
+## 6. Package layout and ownership
+
+Per [ADR 0007](../adr/0007-package-delivery-amends-0004.md) §2 — the dpkg/rpm database is the manifest; no `manifest.txt` ships:
+
+| Artifact | Location | Ownership |
+|---|---|---|
+| Bundled runtime packages | `/opt/fenris/vendor` | package (tree) |
+| Wrapper | `/usr/bin/fenris` | package |
+| Helpers | `/usr/libexec/fenris/{fenris-monitor,fenris-collect}` | package — exactly these two, no new polkit-reachable binaries |
+| Units | `/usr/lib/systemd/system/fenris-collect.{timer,service}` | package (vendor placement; `/etc/systemd/system` is admin-only) |
+| Polkit policy | `/usr/share/polkit-1/actions/com.bongbetic.fenris.monitor.policy` | package |
+| sysusers fragment | `/usr/lib/sysusers.d/fenris.conf` (`g fenris -`) | package |
+| tmpfiles fragment | `/usr/lib/tmpfiles.d/fenris.conf` (`d /var/lib/fenris 2750 root fenris -`) | package |
+| Configuration | `/etc/fenris/fenris.conf` | package as conffile / `%config(noreplace)` — placeholder-commented default, no active selector |
+| Observation store | `/var/lib/fenris/observations.db` (+ WAL, `.bak`) | **never owned, never ghosted** — the package owns the directory only |
+
+## 7. Maintainer-script contracts
+
+- **preinst / %pre:** abort with a pointer to the migration runbook (§9) if `/var/lib/fenris/manifest.txt` **or** `/etc/systemd/system/fenris-collect.timer` exists (dual marker covers pre-manifest make installs). No auto-clean — scripts never delete files outside the package DB.
+- **postinst / %post (install):** `systemd-sysusers`, `systemd-tmpfiles --create`, `systemctl daemon-reload`. Nothing else — no enable, no preset, no start; no preset file ships.
+- **postinst / %post (upgrade):** snapshot `observations.db` → `.bak` (one generation) → forward-only schema migration via target `python3` with `/opt/fenris/vendor` on its import path → `daemon-reload` → restart `fenris-collect.timer` only if unit contents changed **and** it is active. `/var/lib/fenris` is never rebuilt; an in-flight oneshot finishes on its old interpreter.
+- **prerm / %preun:** sanctioned disable (`fenris-monitor disable --now`, closing the monitoring period `user_disabled`) on remove/erase **only, never on upgrade** — deb prerm upgrade case is a no-op; rpm `%preun` gated on `$1 -eq 0`.
+- **Removal mapping:** deb `remove` ≈ `make uninstall` (conffile + store survive); deb `purge` ≈ `make purge` (+ `.bak`, group cleanup); rpm erase ≈ `make uninstall` (unmodified config removed, modified survives as `.rpmsave`); rpm purge = documented manual command.
+
+## 8. Initial configuration
+
+- The device selector is **entered by hand**: root edits `/etc/fenris/fenris.conf` (world-readable, exactly one key per [ADR 0003](../adr/0003-service-lifecycle-and-sanctioned-toggle.md) §3). The shipped default is placeholder-commented and carries no active selector — a fresh install reads as a `configuration error`-free dormant system until the first `fenris monitor resume` + edit, exactly the dormant-install contract.
+- No configuration verb is added to `fenris-monitor`; the polkit surface stays at one binary. (This resolves the open item from [Package ownership + ADR 0004 amendment](https://git.bongbetic.com/xavierk/Fenris/issues/40): nothing in any delivery writes the selector — `make install` never wrote `fenris.conf` either; hand-editing has been the model since ADR 0003.)
+- On upgrade, local edits survive; a changed package default lands as `.dpkg-new` / `.rpmnew`.
+
+## 9. Migration from make-install systems
+
+- **Runbook only** — no migration script, no auto-clean. Population is author machines plus a few testers; store and config survive by path continuity.
+- **Remove-then-install, mandatory:** `sudo make uninstall` (preserves store + `/etc/fenris`) → `apt install fenris` / `dnf install fenris`. **Over-install is forbidden:** stale `/etc/systemd/system/fenris-collect.*` silently shadows vendor units (systemd precedence), `/usr/local/bin/fenris` shadows `/usr/bin/fenris` on PATH.
+- **No-move continuity:** `/var/lib/fenris` untouched; existing `fenris` group → sysusers no-op; existing dir → tmpfiles no-op; hand-written `fenris.conf` survives (dpkg ships the default as `.dpkg-new`; rpm as `.rpmnew`); store schema caught up by the upgrade-path migration.
+- **Reset-to-dormant:** `make uninstall`'s sanctioned disable closes the open period `user_disabled`; after migration the user opts back in with `fenris monitor resume` — one ≤15-min sample gap, honest against the endurance timeline.
+- **Mutual exclusion:** package and `make install` never on the same machine. The dev loop is checkout + `make test`; a dev-local install mode stays in map fog.
+
+## 10. Out of scope
+
+- Building and publishing packages (the follow-up execution effort).
+- Snap/Flatpak/AppImage/Homebrew, PyPI as an install path.
+- Stable/testing channel split, COPR/PPA fallback, arm64 — map fog until demand appears.
+
+---
+
+*Assembled from [Lock channel + toolchain](https://git.bongbetic.com/xavierk/Fenris/issues/38), [Signing + key policy](https://git.bongbetic.com/xavierk/Fenris/issues/39), [Package ownership + ADR 0004 amendment](https://git.bongbetic.com/xavierk/Fenris/issues/40), [Migration path from make-install systems to packages](https://git.bongbetic.com/xavierk/Fenris/issues/41), [Release cadence + stable/testing channel split](https://git.bongbetic.com/xavierk/Fenris/issues/43), [Actions runner availability](https://git.bongbetic.com/xavierk/Fenris/issues/37), and the research tickets [deb + rpm packaging toolchain](https://git.bongbetic.com/xavierk/Fenris/issues/34), [Gitea 1.27 package registry feasibility](https://git.bongbetic.com/xavierk/Fenris/issues/35), [OBS route](https://git.bongbetic.com/xavierk/Fenris/issues/36).*

fenris.sh

@@ -1,128 +0,0 @@
-#!/usr/bin/env bash
-# Fenris — interactive menu for the NVMe wear monitor & dashboard.
-# Created by Bongbetic.
-
-set -euo pipefail
-SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-PY="$SCRIPT_DIR/fenris.py"
-PORT_DEFAULT=8420
-INTERVAL_DEFAULT=300
-
-banner() {
-  cat <<'EOF'
- _____           _
-|   __|___ ___ _| |___
-|   __| -_|   | . |  _|
-|__|  |___|_|_|_|___|_|
-
- NVMe wear monitor & live dashboard
- Created by Bongbetic
-EOF
-}
-
-pause() { read -rp "Press Enter to continue..." _; }
-
-detect_device() {
-  # Query Python's auto-detect for the default device.
-  python3 -c "import sys; sys.path.insert(0,'$SCRIPT_DIR'); from fenris import detect_device; print(detect_device())" 2>/dev/null || echo /dev/nvme0
-}
-
-menu() {
-  clear
-  banner
-  echo
-  echo "  1) Start monitoring (background daemon + dashboard)"
-  echo "  2) Stop monitoring"
-  echo "  3) Status / current wear stats"
-  echo "  4) Take one sample right now"
-  echo "  5) Open dashboard URL"
-  echo "  ---"
-  echo "  h) Help / how this works"
-  echo "  q) Exit"
-  echo
-  read -rp "Choose an option: " choice
-  echo
-  case "$choice" in
-    1) start_flow ;;
-    2) python3 "$PY" stop; pause ;;
-    3) python3 "$PY" status; pause ;;
-    4) read -rp "Device [default: auto-detect]: " dev
-       if [ -z "$dev" ]; then python3 "$PY" sample; else python3 "$PY" sample --device "$dev"; fi
-       pause ;;
-    5) show_url; pause ;;
-    h|H) help_text; pause ;;
-    q|Q) echo "Bye. — Fenris, by Bongbetic"; exit 0 ;;
-    *) echo "Invalid choice."; pause ;;
-  esac
-}
-
-start_flow() {
-  read -rp "NVMe device [Enter = auto-detect]: " dev
-  read -rp "Sample interval in seconds [Enter = ${INTERVAL_DEFAULT}]: " interval
-  read -rp "Dashboard port [Enter = ${PORT_DEFAULT}]: " port
-  interval="${interval:-$INTERVAL_DEFAULT}"
-  port="${port:-$PORT_DEFAULT}"
-
-  args=(start --interval "$interval" --port "$port")
-  if [ -n "${dev:-}" ]; then args+=(--device "$dev"); fi
-
-  echo
-  echo "Note: reading NVMe SMART data needs root."
-  echo "Fenris runs 'sudo -n smartctl ...' (no-prompt sudo). If this fails,"
-  echo "either run this menu with sudo, or allow passwordless smartctl via:"
-  echo "  sudo visudo   ->   youruser ALL=(root) NOPASSWD: /usr/sbin/smartctl"
-  echo
-
-  python3 "$PY" "${args[@]}"
-  pause
-}
-
-show_url() {
-  if [ -f "$SCRIPT_DIR/data/fenris.pid" ]; then
-    # Try to read actual port from running process cmdline, else guess default.
-    local pid port
-    pid=$(<"$SCRIPT_DIR/data/fenris.pid")
-    port=$(tr '\0' '\n' < /proc/"$pid"/cmdline 2>/dev/null | grep -A1 -- '--port' | tail -1 || true)
-    port="${port:-$PORT_DEFAULT}"
-    echo "Dashboard: http://localhost:${port}"
-  else
-    echo "Fenris is not currently running. Start it first (option 1)."
-  fi
-}
-
-help_text() {
-  cat <<EOF
-What Fenris does:
-  - Periodically reads your NVMe drive's SMART health data (via smartctl),
-    including "percentage_used" (the drive's own wear indicator), total
-    bytes written/read, temperature, spare capacity, and error counts.
-  - Logs every sample to: $SCRIPT_DIR/data/history.jsonl
-    and per-hour aggregates to: $SCRIPT_DIR/data/hourly.jsonl (GB/hour,
-    rebuilt from history on restart). Used for the trailing-24h bar chart
-    and exact rolling-24h write volume.
-  - Serves a live HTML dashboard (dense layout, interval-synced polling)
-    with wear-over-time and trailing-24h hourly-write charts, plus a
-    projected life-remaining estimate in hours/days/years derived from
-    your actual rolling-24h write rate and implied TBW endurance.
-
-Requirements:
-  - smartmontools (smartctl) installed.
-  - Root access to read NVMe SMART logs — either run Fenris via sudo,
-    or set up passwordless sudo for smartctl (see option 1).
-
-CLI usage (equivalent to this menu):
-  python3 fenris.py start    [--device /dev/nvme0] [--interval 300] [--port 8420]
-  python3 fenris.py stop
-  python3 fenris.py status
-  python3 fenris.py sample   [--device /dev/nvme0]
-
-Leave it running in the background (option 1) and check back after a
-few days/weeks of normal use — more samples = a more accurate lifespan
-estimate.
-
-Fenris — created by Bongbetic.
-EOF
-}
-
-# Loop instead of recurse to avoid stack overflow.
-while true; do menu; done

packaging/fenris.conf

@@ -0,0 +1,16 @@
+# Fenris configuration
+#
+# This file is managed by the fenris package. Local edits are preserved
+# across upgrades; changed defaults appear as .dpkg-new / .rpmnew.
+#
+# The device selector specifies which NVMe drive to monitor.
+# Uncomment and set exactly one device path:
+#
+#   device = /dev/disk/by-id/nvme-Samsung_SSD_980_PRO_500GB_S5PANS0T123456
+#
+# The observation store path is optional and defaults to
+# /var/lib/fenris/observations.db when unset:
+#
+#   store_path = /var/lib/fenris/observations.db
+#
+# See https://git.bongbetic.com/xavierk/Fenris for documentation.

packaging/fenris.repo

@@ -0,0 +1,7 @@
+[fenris]
+name=Fenris NVMe Monitor
+baseurl=https://git.bongbetic.com/api/packages/xavierk/rpm/fenris
+enabled=1
+gpgcheck=1
+gpgkey=https://git.bongbetic.com/xavierk/Fenris/raw/branch/main/packaging/keys/fenris-packaging.asc
+repo_gpgcheck=0

packaging/keys/fenris-packaging.asc

@@ -0,0 +1,40 @@
+# Fenris Packaging Key
+#
+# Public half of dedicated RSA-3072 key used to sign RPM payloads and
+# clearsign SHA256SUMS manifests.
+#
+# Fingerprint: CE4542E1E23EB50F09EDFFA5A5E8B22D1872FB07
+# Algorithm: RSA 3072
+# UID: Fenris Packaging <packaging@bongbetic.com>
+# Expiry: 2 years from creation
+#
+# Raw URL:
+#   https://git.bongbetic.com/xavierk/Fenris/raw/branch/main/packaging/keys/fenris-packaging.asc
+#
+# The private half lives in approved secret storage only. Each release uses
+# import -> sign -> delete. See docs/install/signing-key-ceremony.md.
+#
+-----BEGIN PGP PUBLIC KEY BLOCK-----
+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+=V4/R
+-----END PGP PUBLIC KEY BLOCK-----

packaging/nfpm.yaml

@@ -0,0 +1,60 @@
+name: fenris
+arch: amd64
+platform: linux
+version: "${VERSION}"
+maintainer: Fenris Maintainers <ops@bongbetic.com>
+description: >
+  NVMe wear monitor with persistent TUI — observes real-world drive use and
+  translates it into an understandable endurance outlook.
+homepage: https://git.bongbetic.com/xavierk/Fenris
+license: Proprietary
+
+depends:
+  - python3 (>= 3.10)
+  - smartmontools
+  - systemd
+
+contents:
+  # Staged tree: runtime packages, wrapper, helpers, units, polkit, sysusers, tmpfiles
+  - src: build/stage/
+    dst: /
+    type: tree
+
+  # Configuration directory
+  - dst: /etc/fenris
+    type: dir
+    file_info:
+      mode: 0755
+
+  # Default placeholder-commented config (deb conffile / rpm %config(noreplace))
+  - src: packaging/fenris.conf
+    dst: /etc/fenris/fenris.conf
+    type: config|noreplace
+    file_info:
+      mode: 0644
+
+  # Observation store directory — owned by package, never packed.
+  # Store files (observations.db, WAL sidecars, .bak) are never owned.
+  - dst: /var/lib/fenris
+    type: dir
+    file_info:
+      mode: 2750
+      group: fenris
+
+scripts:
+  preinstall: packaging/preinst.sh
+  postinstall: packaging/postinst.sh
+  preremove: packaging/prerm.sh
+  postremove: packaging/postrm.sh
+
+overrides:
+  rpm:
+    depends:
+      - python3 >= 3.10
+      - smartmontools
+      - systemd
+    scripts:
+      preinstall: packaging/preinst.sh
+      postinstall: packaging/rpm/post.sh
+      preremove: packaging/rpm/preun.sh
+      postremove: packaging/rpm/postun.sh

packaging/postinst.sh

@@ -0,0 +1,56 @@
+#!/bin/sh
+# postinst — deb install and upgrade paths (spec §7).
+#
+# dpkg calls: postinst configure [most-recently-configured-version]
+#   fresh install: $1 = "configure", $2 = ""
+#   upgrade:       $1 = "configure", $2 = old version
+set -eu
+
+STORE_DIR="/var/lib/fenris"
+STORE_DB="${STORE_DIR}/observations.db"
+STORE_BAK="${STORE_DIR}/observations.db.bak"
+RUNTIME_PYTHON="/usr/bin/python3"
+VENDOR_DIR="/opt/fenris/vendor"
+
+case "${1:-}" in
+    configure)
+        if [ -n "${2:-}" ]; then
+            # Upgrade — snapshot, migration, daemon-reload, conditional timer restart
+            if [ -f "${STORE_DB}" ]; then
+                cp "${STORE_DB}" "${STORE_BAK}" 2>/dev/null || true
+            fi
+            if [ -d "${VENDOR_DIR}" ] && [ -f "${STORE_DB}" ]; then
+                PYTHONPATH="${VENDOR_DIR}" "${RUNTIME_PYTHON}" -c "
+from fenris.store import migrate_to_latest
+from pathlib import Path
+n = migrate_to_latest(Path('${STORE_DB}'))
+print(f'Fenris migration: {n} step(s) applied') if n else None
+" 2>&1 || echo "Fenris: migration skipped (store not yet initialized)"
+            fi
+            # Capture running unit content BEFORE daemon-reload (spec §7)
+            RUNNING_UNITS=""
+            for unit in fenris-collect.timer; do
+                if systemctl is-active --quiet "${unit}" 2>/dev/null; then
+                    RUNNING_UNITS="${RUNNING_UNITS} ${unit}"
+                fi
+            done
+            systemctl daemon-reload 2>/dev/null || true
+            # Restart timer only if unit contents changed AND active
+            for unit in ${RUNNING_UNITS}; do
+                OLD_CONTENT="$(mktemp)"
+                NEW_PATH="/usr/lib/systemd/system/${unit}"
+                systemctl cat "${unit}" > "${OLD_CONTENT}" 2>/dev/null || true
+                if ! diff -q "${OLD_CONTENT}" "${NEW_PATH}" > /dev/null 2>&1; then
+                    systemctl restart "${unit}" 2>/dev/null || true
+                fi
+                rm -f "${OLD_CONTENT}"
+            done
+        fi
+        # sysusers, tmpfiles, daemon-reload (both fresh install and upgrade)
+        systemd-sysusers  || true
+        systemd-tmpfiles --create  || true
+        systemctl daemon-reload  || true
+        ;;
+    abort-upgrade|abort-install|disappear)
+        ;;
+esac

packaging/postrm.sh

@@ -0,0 +1,19 @@
+#!/bin/sh
+# postrm — deb post-removal (spec §7, §9).
+#
+# dpkg calls: postrm remove    (after package files removed)
+#             postrm purge     (after conffiles and config removed)
+set -eu
+
+case "${1:-}" in
+    purge)
+        rm -rf /etc/fenris
+        rm -rf /var/lib/fenris
+        if getent group fenris > /dev/null 2>&1; then
+            groupdel fenris 2>/dev/null || true
+        fi
+        ;;
+    remove|upgrade|failed-upgrade|abort-install|abort-upgrade|disappear)
+        ;;
+esac
+systemctl daemon-reload 2>/dev/null || true

packaging/preinst.sh

@@ -0,0 +1,19 @@
+#!/bin/sh
+# preinst — abort if make-install remnants detected (spec §7, §9).
+set -eu
+
+MARKER1="/var/lib/fenris/manifest.txt"
+MARKER2="/etc/systemd/system/fenris-collect.timer"
+
+if [ -f "${MARKER1}" ] || [ -f "${MARKER2}" ]; then
+    echo >&2
+    echo >&2 "Fenris make-install remnants detected — refusing to install."
+    echo >&2
+    echo >&2 "Migrate to the package with:"
+    echo >&2 "  sudo make uninstall   # removes make-install files, preserves store + config"
+    echo >&2 "  sudo apt install fenris   # or: sudo dnf install fenris"
+    echo >&2
+    echo >&2 "See: https://git.bongbetic.com/xavierk/Fenris/blob/main/docs/spec/release-packaging.md#9-migration-from-make-install-systems"
+    echo >&2
+    exit 1
+fi

packaging/prerm.sh

@@ -0,0 +1,20 @@
+#!/bin/sh
+# prerm — deb pre-removal (spec §7).
+#
+# dpkg calls: prerm remove  (package being removed)
+#             prerm upgrade  (old version about to be replaced)
+set -eu
+
+case "${1:-}" in
+    remove)
+        # Sanctioned disable — close monitoring period (spec §7)
+        if [ -x /usr/libexec/fenris/fenris-monitor ]; then
+            /usr/libexec/fenris/fenris-monitor disable --now 2>/dev/null || true
+        fi
+        systemctl stop fenris-collect.timer 2>/dev/null || true
+        systemctl disable fenris-collect.timer 2>/dev/null || true
+        ;;
+    upgrade)
+        # Never interrupt monitoring on upgrade
+        ;;
+esac

packaging/release-footer.md

@@ -0,0 +1,20 @@
+## Install
+
+Install Fenris from its package channel after following the [package setup instructions](https://git.bongbetic.com/xavierk/Fenris/src/branch/main/README.md#install-from-package-recommended):
+
+```bash
+sudo apt update && sudo apt install fenris       # Debian / Ubuntu
+sudo dnf install fenris                          # Fedora
+sudo zypper install fenris                       # openSUSE Tumbleweed
+```
+
+## Verify downloads
+
+```bash
+gpg --output SHA256SUMS --decrypt SHA256SUMS.asc
+sha256sum -c SHA256SUMS
+```
+
+## Rollback
+
+Installing an older package over a newer observation store is unsupported. Restore the observation-store snapshot, then install the earlier Release; see the [upgrade and rollback guidance](https://git.bongbetic.com/xavierk/Fenris/src/branch/main/README.md#upgrade).

packaging/rpm/post.sh

@@ -0,0 +1,49 @@
+#!/bin/sh
+# RPM %post — post-install/upgrade scriptlet (spec §7).
+set -eu
+
+STORE_DIR="/var/lib/fenris"
+STORE_DB="${STORE_DIR}/observations.db"
+STORE_BAK="${STORE_DIR}/observations.db.bak"
+RUNTIME_PYTHON="/usr/bin/python3"
+VENDOR_DIR="/opt/fenris/vendor"
+
+if [ "$1" -eq 1 ]; then
+    # Fresh install
+    systemd-sysusers  || true
+    systemd-tmpfiles --create  || true
+    systemctl daemon-reload  || true
+elif [ "$1" -ge 2 ]; then
+    # Upgrade — snapshot, migration, daemon-reload, conditional timer restart
+    if [ -f "${STORE_DB}" ]; then
+        cp "${STORE_DB}" "${STORE_BAK}" 2>/dev/null || true
+    fi
+    if [ -d "${VENDOR_DIR}" ] && [ -f "${STORE_DB}" ]; then
+        PYTHONPATH="${VENDOR_DIR}" "${RUNTIME_PYTHON}" -c "
+from fenris.store import migrate_to_latest
+from pathlib import Path
+n = migrate_to_latest(Path('${STORE_DB}'))
+print(f'Fenris migration: {n} step(s) applied') if n else None
+" 2>&1 || echo "Fenris: migration skipped (store not yet initialized)"
+    fi
+    # Capture running unit content BEFORE daemon-reload (spec §7)
+    RUNNING_UNITS=""
+    for unit in fenris-collect.timer; do
+        if systemctl is-active --quiet "${unit}" 2>/dev/null; then
+            RUNNING_UNITS="${RUNNING_UNITS} ${unit}"
+        fi
+    done
+    systemctl daemon-reload 2>/dev/null || true
+    # Restart timer only if unit contents changed AND active
+    for unit in ${RUNNING_UNITS}; do
+        OLD_CONTENT="$(mktemp)"
+        NEW_PATH="/usr/lib/systemd/system/${unit}"
+        systemctl cat "${unit}" > "${OLD_CONTENT}" 2>/dev/null || true
+        if ! diff -q "${OLD_CONTENT}" "${NEW_PATH}" > /dev/null 2>&1; then
+            systemctl restart "${unit}" 2>/dev/null || true
+        fi
+        rm -f "${OLD_CONTENT}"
+    done
+    # Re-apply placement modes (store dir group access, issue #54)
+    systemd-tmpfiles --create  || true
+fi

packaging/rpm/postun.sh

@@ -0,0 +1,13 @@
+#!/bin/sh
+# RPM %postun — post-uninstall scriptlet (spec §7).
+set -eu
+
+if [ "$1" -eq 0 ]; then
+    # Package fully erased — remove config, store, group
+    rm -rf /etc/fenris
+    rm -rf /var/lib/fenris
+    if getent group fenris > /dev/null 2>&1; then
+        groupdel fenris 2>/dev/null || true
+    fi
+fi
+systemctl daemon-reload 2>/dev/null || true

packaging/rpm/preun.sh

@@ -0,0 +1,13 @@
+#!/bin/sh
+# RPM %preun — pre-uninstall scriptlet (spec §7).
+set -eu
+
+if [ "$1" -eq 0 ]; then
+    # Package is being erased — sanctioned disable (spec §7)
+    if [ -x /usr/libexec/fenris/fenris-monitor ]; then
+        /usr/libexec/fenris/fenris-monitor disable --now 2>/dev/null || true
+    fi
+    systemctl stop fenris-collect.timer 2>/dev/null || true
+    systemctl disable fenris-collect.timer 2>/dev/null || true
+fi
+# On upgrade ($1 -ge 1): do nothing

packaging/stage.sh

@@ -0,0 +1,91 @@
+#!/usr/bin/env bash
+# Stage a packaging tree at build/stage/ for nfpm consumption.
+#
+# Usage: packaging/stage.sh [VERSION]
+#
+# VERSION defaults to the version in pyproject.toml.
+# The staged tree contains:
+#   /opt/fenris/vendor/   — bundled pure-Python application dependencies
+#   /usr/bin/fenris       — unprivileged wrapper
+#   /usr/libexec/fenris/  — fenris-monitor, fenris-collect
+#   /usr/lib/systemd/system/ — fenris-collect.{timer,service}
+#   /usr/share/polkit-1/actions/ — polkit policy
+#   /usr/lib/sysusers.d/fenris.conf
+#   /usr/lib/tmpfiles.d/fenris.conf
+set -euo pipefail
+
+REPO_ROOT="$(cd "$(dirname "$0")/.." && pwd)"
+STAGE_DIR="${REPO_ROOT}/build/stage"
+
+# --- Resolve version ---
+if [ -n "${1:-}" ]; then
+    VERSION="$1"
+else
+    VERSION="$(sed -n 's/^version = "\(.*\)"/\1/p' "${REPO_ROOT}/pyproject.toml")"
+fi
+
+if [ -z "${VERSION}" ]; then
+    echo "Error: could not determine version" >&2
+    exit 1
+fi
+echo "Staging fenris ${VERSION} ..."
+
+# --- Clean previous stage ---
+rm -rf "${STAGE_DIR}"
+mkdir -p "${STAGE_DIR}"
+
+# --- Use pre-built wheel from dist/ ---
+WHEEL=$(ls "${REPO_ROOT}"/dist/fenris-"${VERSION}"-*.whl 2>/dev/null | head -1)
+if [ -z "${WHEEL}" ]; then
+    echo "Error: no wheel found in dist/ — run 'make dist/fenris-*.whl' first" >&2
+    exit 1
+fi
+echo "  Using wheel: $(basename "${WHEEL}")"
+
+# --- Vendor runtime packages without an interpreter ---
+# A copied Python binary contains an ABI and build-host dynamic-library path.
+# It fails after rolling-distribution Python upgrades (for example Tumbleweed
+# 3.12 -> 3.13).  Fenris and its locked dependencies are pure Python, so place
+# them in a version-neutral directory and execute with the target's python3.
+echo "  Installing version-neutral runtime packages ..."
+VENDOR_DIR="${STAGE_DIR}/opt/fenris/vendor"
+mkdir -p "${VENDOR_DIR}"
+python3 -m pip install --disable-pip-version-check --no-compile \
+    --target "${VENDOR_DIR}" -r "${REPO_ROOT}/requirements.txt" "${WHEEL}"
+
+# --- Inject version into wrapper from pyproject.toml ---
+# The wrapper has a hardcoded version string; patch it for packaging.
+WRAPPER_SRC="${REPO_ROOT}/scripts/fenris"
+WRAPPER_DST="${STAGE_DIR}/usr/bin/fenris"
+mkdir -p "$(dirname "${WRAPPER_DST}")"
+sed "s|version=\"%(prog)s [0-9.]*\"|version=\"%(prog)s ${VERSION}\"|g" \
+    "${WRAPPER_SRC}" > "${WRAPPER_DST}"
+chmod 0755 "${WRAPPER_DST}"
+
+# --- Privileged helpers ---
+echo "  Installing helpers ..."
+mkdir -p "${STAGE_DIR}/usr/libexec/fenris"
+install -m 0755 "${REPO_ROOT}/src/fenris/monitor.py" "${STAGE_DIR}/usr/libexec/fenris/fenris-monitor"
+install -m 0755 "${REPO_ROOT}/src/fenris/collect.py" "${STAGE_DIR}/usr/libexec/fenris/fenris-collect"
+
+# --- systemd units (vendor placement) ---
+echo "  Installing systemd units ..."
+mkdir -p "${STAGE_DIR}/usr/lib/systemd/system"
+install -m 0644 "${REPO_ROOT}/units/fenris-collect.timer"   "${STAGE_DIR}/usr/lib/systemd/system/"
+install -m 0644 "${REPO_ROOT}/units/fenris-collect.service" "${STAGE_DIR}/usr/lib/systemd/system/"
+
+# --- polkit policy ---
+echo "  Installing polkit policy ..."
+mkdir -p "${STAGE_DIR}/usr/share/polkit-1/actions"
+install -m 0644 "${REPO_ROOT}/polkit/com.bongbetic.fenris.monitor.policy" \
+    "${STAGE_DIR}/usr/share/polkit-1/actions/"
+
+# --- sysusers and tmpfiles fragments ---
+echo "  Installing sysusers/tmpfiles fragments ..."
+mkdir -p "${STAGE_DIR}/usr/lib/sysusers.d"
+install -m 0644 "${REPO_ROOT}/packaging/sysusers.d/fenris.conf" "${STAGE_DIR}/usr/lib/sysusers.d/"
+
+mkdir -p "${STAGE_DIR}/usr/lib/tmpfiles.d"
+install -m 0644 "${REPO_ROOT}/packaging/tmpfiles.d/fenris.conf" "${STAGE_DIR}/usr/lib/tmpfiles.d/"
+
+echo "Stage complete: ${STAGE_DIR}"

packaging/sysusers.d/fenris.conf

@@ -0,0 +1,3 @@
+# System user/group for Fenris observation store access
+# Created by systemd-sysusers during package install
+g fenris -

packaging/tmpfiles.d/fenris.conf

@@ -0,0 +1,2 @@
+# Type  Path              Mode  User   Group  Age  Argument
+d       /var/lib/fenris   2770  root   fenris  -   -

polkit/com.bongbetic.fenris.monitor.policy

@@ -0,0 +1,21 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE policyconfig PUBLIC
+ "-//freedesktop//DTD PolicyKit Policy Configuration 1.0//EN"
+ "http://www.freedesktop.org/standards/PolicyKit/1/policyconfig.dtd">
+<policyconfig>
+  <vendor>bongbetic</vendor>
+  <vendor_url>https://bongbetic.com</vendor_url>
+
+  <action id="com.bongbetic.fenris.monitor">
+    <description>Fenris Monitor Helper</description>
+    <message>Authentication is required to manage Fenris monitoring.</message>
+
+    <defaults>
+      <allow_any>no</allow_any>
+      <allow_inactive>no</allow_inactive>
+      <allow_active>auth_admin</allow_active>
+    </defaults>
+
+    <annotate key="org.freedesktop.policykit.imply">org.freedesktop.systemd1.manage-units</annotate>
+  </action>
+</policyconfig>

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "fenris"
-version = "0.3.0"
+version = "0.3.5"
 description = "NVMe wear monitor with persistent TUI"
 requires-python = ">=3.9"
 dependencies = [
@@ -11,6 +11,7 @@ dependencies = [
 dev = [
     "pytest>=7.0.0",
     "pytest-cov>=4.0.0",
+    "pytest-asyncio>=0.20.0",
 ]
 
 [tool.pytest.ini_options]

requirements.txt

@@ -0,0 +1,11 @@
+# Fenris dependency lockfile
+# Exact pins for reproducible installs (IN-8)
+# Refresh with: make update-deps
+textual==8.2.8
+rich==15.0.0
+markdown-it-py==4.2.0
+mdit-py-plugins==0.6.1
+mdurl==0.1.2
+platformdirs==4.11.7
+Pygments==2.21.0
+linkify-it-py==2.2.0

scripts/extract_changelog.py

@@ -0,0 +1,99 @@
+#!/usr/bin/env python3
+"""Extract one validated Keep a Changelog version section."""
+from __future__ import annotations
+
+import argparse
+from datetime import date
+from pathlib import Path
+import re
+import sys
+
+
+class ChangelogError(ValueError):
+    """A release cannot safely use the supplied changelog."""
+
+
+_SEMVER = r"(?:0|[1-9]\d*)\.(?:0|[1-9]\d*)\.(?:0|[1-9]\d*)"
+_VERSION_HEADING = re.compile(
+    rf"^## \[(?P<version>{_SEMVER})\] - (?P<date>.+)$", re.MULTILINE
+)
+
+
+def extract_version_section(changelog: str, version: str) -> str:
+    """Return *version*'s changelog section without altering its bytes.
+
+    The section ends immediately before the next level-two heading.  A release
+    cannot use an absent, empty, or malformed version section.
+    """
+    if not re.fullmatch(_SEMVER, version):
+        raise ChangelogError(f"requested version is not bare semver: {version!r}")
+
+    heading = next(
+        (match for match in _VERSION_HEADING.finditer(changelog)
+         if match.group("version") == version),
+        None,
+    )
+    if heading is None:
+        if re.search(rf"^## \[{re.escape(version)}\].*$", changelog, re.MULTILINE):
+            raise ChangelogError(f"version {version} has a malformed heading or date")
+        raise ChangelogError(f"version {version} is missing from the changelog")
+
+    heading_date = heading.group("date")
+    if not re.fullmatch(r"\d{4}-\d{2}-\d{2}", heading_date):
+        raise ChangelogError(f"version {version} has a malformed release date")
+    try:
+        date.fromisoformat(heading_date)
+    except ValueError as error:
+        raise ChangelogError(f"version {version} has a malformed release date") from error
+
+    next_heading = re.search(r"^## ", changelog[heading.end():], re.MULTILINE)
+    section_end = heading.end() + next_heading.start() if next_heading else len(changelog)
+    section = changelog[heading.start():section_end]
+    if not re.search(r"^- \S", section[heading.end() - heading.start():], re.MULTILINE):
+        raise ChangelogError(f"version {version} has an empty changelog section")
+    return section
+
+
+def extract_changelog(path: Path, version: str) -> str:
+    """Read and extract a requested version from a changelog file."""
+    try:
+        return extract_version_section(path.read_text(encoding="utf-8"), version)
+    except OSError as error:
+        raise ChangelogError(f"cannot read changelog {path}: {error.strerror}") from error
+
+
+def assemble_release_body(section: str, footer: str) -> str:
+    """Append standing guidance while preserving the extracted section verbatim."""
+    separator = "\n" if section.endswith("\n") else "\n\n"
+    return f"{section}{separator}{footer}"
+
+
+def main(argv: list[str] | None = None) -> int:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("changelog", type=Path)
+    parser.add_argument("version")
+    parser.add_argument(
+        "--footer",
+        type=Path,
+        help="append this standing release guidance after the extracted section",
+    )
+    args = parser.parse_args(argv)
+    try:
+        section = extract_changelog(args.changelog, args.version)
+        if args.footer:
+            try:
+                footer = args.footer.read_text(encoding="utf-8")
+            except OSError as error:
+                raise ChangelogError(
+                    f"cannot read release footer {args.footer}: {error.strerror}"
+                ) from error
+            section = assemble_release_body(section, footer)
+        sys.stdout.write(section)
+    except ChangelogError as error:
+        print(f"::error::{error}", file=sys.stderr)
+        return 1
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

scripts/fenris

@@ -0,0 +1,227 @@
+#!/usr/bin/env python3
+"""fenris: unprivileged entry point for the Fenris TUI and CLI.
+
+With no arguments, opens the TUI.
+Subcommands route through fenris-monitor for privileged operations.
+
+Spec: §1.2, §8.4
+"""
+import argparse
+import os
+import subprocess
+import sys
+from pathlib import Path
+
+
+def add_runtime_packages() -> None:
+    """Make the package-owned, pure-Python dependencies importable.
+
+    RPM and deb installations deliberately use the target system's Python.
+    Their dependencies are vendored without a copied interpreter so a distro
+    Python minor-version update cannot leave Fenris linked to a removed ABI.
+    The legacy development install keeps its venv fallback.
+    """
+    runtime_dir = Path("/opt/fenris")
+    vendor_dir = runtime_dir / "vendor"
+    if vendor_dir.is_dir():
+        sys.path.insert(0, str(vendor_dir))
+        return
+
+    site_packages = next((runtime_dir / "lib").glob("python*/site-packages"), None)
+    if site_packages:
+        sys.path.insert(0, str(site_packages))
+
+
+add_runtime_packages()
+
+
+def is_root() -> bool:
+    """Check if running as root."""
+    return os.geteuid() == 0
+
+
+def run_monitor(*args: str) -> None:
+    """Run fenris-monitor with the given arguments.
+
+    If not root, re-exec under pkexec.
+    """
+    monitor_cmd = "/usr/libexec/fenris/fenris-monitor"
+
+    if is_root():
+        result = subprocess.run([monitor_cmd] + list(args))
+        sys.exit(result.returncode)
+    else:
+        # Use pkexec to elevate
+        pkexec = subprocess.run(
+            ["which", "pkexec"], capture_output=True
+        )
+        if pkexec.returncode != 0:
+            print(
+                "Error: No polkit agent available. "
+                "Run as root: sudo fenris-monitor ...",
+                file=sys.stderr,
+            )
+            sys.exit(1)
+        result = subprocess.run(["pkexec", monitor_cmd] + list(args))
+        sys.exit(result.returncode)
+
+
+def cmd_tui(args: argparse.Namespace) -> None:
+    """Open the TUI."""
+    from fenris.tui import run_tui
+    run_tui()
+
+
+def cmd_status(args: argparse.Namespace) -> None:
+    """Show status."""
+    from fenris.status import render_status
+    print(render_status())
+
+
+def cmd_sample(args: argparse.Namespace) -> None:
+    """Trigger on-demand collection."""
+    run_monitor("collect")
+
+
+def cmd_monitor_pause(args: argparse.Namespace) -> None:
+    """Pause monitoring."""
+    # Pause asks confirmation (§7.4)
+    if not args.yes:
+        response = input("Pause monitoring? [y/N] ")
+        if response.lower() not in ("y", "yes"):
+            print("Aborted.")
+            return
+
+    run_monitor("disable", "--now")
+
+
+def cmd_monitor_resume(args: argparse.Namespace) -> None:
+    """Resume monitoring."""
+    # Resume does not ask confirmation (§7.4)
+    run_monitor("enable", "--now")
+
+
+def cmd_baseline_set(args: argparse.Namespace) -> None:
+    """Set baseline."""
+    run_monitor("baseline", "set", args.baseline_json)
+
+
+def cmd_baseline_clear(args: argparse.Namespace) -> None:
+    """Clear baseline."""
+    run_monitor("baseline", "clear")
+
+
+def cmd_import(args: argparse.Namespace) -> None:
+    """Import legacy history."""
+    # This is a one-off migration, not a privileged operation
+    print("Legacy import: use fenris-import directly")
+
+
+def cmd_migrate(args: argparse.Namespace) -> None:
+    """Apply forward-only schema migrations (IN-5, IN-6).
+    
+    Called by 'sudo make upgrade'. Raises on newer-schema store.
+    """
+    from fenris.store import migrate_to_latest
+    from pathlib import Path
+
+    store_path = Path("/var/lib/fenris/observations.db")
+    if not store_path.exists():
+        print("No observation store found — nothing to migrate.")
+        return
+
+    steps = migrate_to_latest(store_path)
+    if steps:
+        print(f"Migration complete: {steps} step(s) applied.")
+    else:
+        print("Schema already current.")
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(
+        prog="fenris",
+        description="Fenris NVMe endurance monitor",
+    )
+    parser.add_argument(
+        "--version", action="version", version="%(prog)s 0.3.0"
+    )
+
+    subparsers = parser.add_subparsers(dest="command")
+
+    # Default: TUI (no subcommand)
+    subparsers.add_parser("tui", help="Open the TUI (default)")
+
+    # Status
+    subparsers.add_parser("status", help="Show status")
+
+    # Sample (on-demand collection)
+    subparsers.add_parser("sample", help="Trigger on-demand collection")
+
+    # Monitor subcommand
+    monitor_parser = subparsers.add_parser("monitor", help="Monitor control")
+    monitor_sub = monitor_parser.add_subparsers(dest="monitor_action")
+
+    # monitor pause
+    pause_parser = monitor_sub.add_parser("pause", help="Pause monitoring")
+    pause_parser.add_argument(
+        "-y", "--yes", action="store_true", help="Skip confirmation"
+    )
+    pause_parser.set_defaults(func=cmd_monitor_pause)
+
+    # monitor resume
+    resume_parser = monitor_sub.add_parser("resume", help="Resume monitoring")
+    resume_parser.set_defaults(func=cmd_monitor_resume)
+
+    # Baseline subcommand
+    baseline_parser = subparsers.add_parser("baseline", help="Baseline operations")
+    baseline_sub = baseline_parser.add_subparsers(dest="baseline_action")
+
+    baseline_set = baseline_sub.add_parser("set", help="Set baseline")
+    baseline_set.add_argument("baseline_json", help="Baseline JSON data")
+    baseline_set.set_defaults(func=cmd_baseline_set)
+
+    baseline_clear = baseline_sub.add_parser("clear", help="Clear baseline")
+    baseline_clear.set_defaults(func=cmd_baseline_clear)
+
+    # Import
+    import_parser = subparsers.add_parser("import", help="Import legacy history")
+    import_parser.add_argument("path", help="Path to history.jsonl")
+    import_parser.set_defaults(func=cmd_import)
+
+    # Migrate (IN-5, IN-6) — called by upgrade, not for human use
+    migrate_parser = subparsers.add_parser("migrate", help=argparse.SUPPRESS)
+    migrate_parser.set_defaults(func=cmd_migrate)
+
+    # Rejected commands
+    for cmd in ["start", "stop", "run"]:
+        reject_parser = subparsers.add_parser(cmd, help=argparse.SUPPRESS)
+        reject_parser.set_defaults(func=lambda a: print(
+            f"'{cmd}' is not a valid command. "
+            f"Use 'fenris monitor resume' instead.",
+            file=sys.stderr,
+        ))
+
+    args = parser.parse_args()
+
+    if args.command is None or args.command == "tui":
+        cmd_tui(args)
+    elif args.command == "status":
+        cmd_status(args)
+    elif args.command == "sample":
+        cmd_sample(args)
+    elif args.command == "monitor":
+        if args.monitor_action is None:
+            monitor_parser.error("a subcommand is required")
+        args.func(args)
+    elif args.command == "baseline":
+        if args.baseline_action is None:
+            baseline_parser.error("a subcommand is required")
+        args.func(args)
+    elif args.command == "import":
+        cmd_import(args)
+    elif args.command == "migrate":
+        cmd_migrate(args)
+
+
+if __name__ == "__main__":
+    main()

scripts/release.sh

@@ -0,0 +1,206 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# Fenris one-command release flow (issue #52).
+# Builds both packages, signs, uploads to registry, creates release entry,
+# and attaches artifacts — or in dry-run mode, prints every command.
+#
+# Usage:
+#   scripts/release.sh --dry-run          # Print commands without executing
+#   scripts/release.sh --publish          # Execute the full release flow
+#
+# Environment:
+#   GITEA_TOKEN   - API token for Gitea registry and release API
+#   PACKAGING_KEY - GPG key UID (default: packaging@bongbetic.com)
+#
+# Spec: release-packaging.md §5
+
+# ── Defaults ─────────────────────────────────────────────────────────────
+
+DRY_RUN=false
+PUBLISH=false
+GITEA_URL="https://git.bongbetic.com"
+GITEA_OWNER="xavierk"
+GITEA_REPO="Fenris"
+PACKAGING_KEY="${PACKAGING_KEY:-packaging@bongbetic.com}"
+
+CODENAMES=(bookworm jammy noble)
+RPM_GROUP="fenris"
+
+# ── Parse arguments ──────────────────────────────────────────────────────
+
+for arg in "$@"; do
+    case "$arg" in
+        --dry-run) DRY_RUN=true ;;
+        --publish) PUBLISH=true ;;
+        --help|-h)
+            echo "Usage: $0 [--dry-run | --publish]"
+            echo ""
+            echo "Modes:"
+            echo "  --dry-run   Print commands without executing (default)"
+            echo "  --publish   Execute the full release flow"
+            echo ""
+            echo "Environment:"
+            echo "  GITEA_TOKEN   API token for Gitea registry and release API"
+            echo "  PACKAGING_KEY GPG key UID (default: packaging@bongbetic.com)"
+            exit 0
+            ;;
+        *)
+            echo "Unknown argument: $arg" >&2
+            echo "Usage: $0 [--dry-run | --publish]" >&2
+            exit 1
+            ;;
+    esac
+done
+
+if ! $DRY_RUN && ! $PUBLISH; then
+    DRY_RUN=true
+fi
+
+# ── Helpers ──────────────────────────────────────────────────────────────
+
+_version() {
+    sed -n 's/^version = "\(.*\)"/\1/p' pyproject.toml
+}
+
+_deb_name() {
+    local ver="$1"
+    echo "fenris_${ver}_amd64.deb"
+}
+
+_rpm_name() {
+    local ver="$1" rel="$2"
+    echo "fenris-${ver}-${rel}.x86_64.rpm"
+}
+
+_run() {
+    if $DRY_RUN; then
+        echo "  $*"
+    else
+        eval "$@"
+    fi
+}
+
+# ── Main ─────────────────────────────────────────────────────────────────
+
+VERSION=$(_version)
+REVISION=1
+DEB=$(_deb_name "$VERSION")
+RPM=$(_rpm_name "$VERSION" "$REVISION")
+
+echo "=== Fenris Release v${VERSION} ==="
+echo ""
+
+if $DRY_RUN; then
+    echo "[dry-run] Commands below will be executed in --publish mode."
+    echo ""
+fi
+
+# ── Step 1: Build both formats ──────────────────────────────────────────
+
+echo "--- Build packages ---"
+_run "make package"
+echo ""
+
+# ── Step 2: Sign RPM payload ────────────────────────────────────────────
+
+echo "--- Sign RPM payload ---"
+_run "rpmsign --addsign --define '_gpg_name ${PACKAGING_KEY}' dist/${RPM}"
+echo ""
+
+# ── Step 3: Generate and clearsign SHA256SUMS ────────────────────────────
+
+echo "--- Generate SHA256SUMS ---"
+_run "cd dist && sha256sum ${DEB} ${RPM} > SHA256SUMS"
+echo ""
+
+echo "--- Clearsign SHA256SUMS ---"
+_run "gpg --batch --yes --clearsign --local-user ${PACKAGING_KEY} dist/SHA256SUMS"
+echo ""
+
+# ── Step 4: Upload to Gitea package registry ─────────────────────────────
+
+echo "--- Upload packages to registry ---"
+for codename in "${CODENAMES[@]}"; do
+    _run "curl --fail -X PUT -u ${GITEA_OWNER}:\$GITEA_TOKEN -T dist/${DEB} '${GITEA_URL}/api/packages/${GITEA_OWNER}/debian/pool/${codename}/main/upload'"
+done
+_run "curl --fail -X PUT -u ${GITEA_OWNER}:\$GITEA_TOKEN -T dist/${RPM} '${GITEA_URL}/api/packages/${GITEA_OWNER}/rpm/${RPM_GROUP}/upload'"
+echo ""
+
+# ── Step 5: Create Gitea release with notes ─────────────────────────────
+
+echo "--- Create Gitea release ---"
+_release_notes="Release v${VERSION}
+
+## Packages
+
+Install via apt (Debian/Ubuntu):
+
+\`\`\`bash
+curl --fail -fsSL https://git.bongbetic.com/${GITEA_OWNER}/${GITEA_REPO}/raw/branch/main/packaging/keys/fenris-packaging.asc | sudo gpg --dearmor -o /etc/apt/keyrings/fenris.asc
+echo \"deb [signed-by=/etc/apt/keyrings/fenris.asc] https://git.bongbetic.com/api/packages/${GITEA_OWNER}/debian bookworm main\" | sudo tee /etc/apt/sources.list.d/fenris.list
+sudo apt update && sudo apt install fenris
+\`\`\`
+
+Install via dnf (Fedora):
+
+\`\`\`bash
+sudo dnf config-manager --add-repo https://git.bongbetic.com/${GITEA_OWNER}/${GITEA_REPO}/raw/branch/main/packaging/fenris.repo
+sudo dnf install fenris
+\`\`\`
+
+## Verification
+
+\`\`\`bash
+rpm -Kv fenris-${VERSION}-1.x86_64.rpm
+gpg --verify SHA256SUMS.asc SHA256SUMS
+\`\`\`
+
+## Artifacts
+
+- \`dist/${DEB}\` (Debian/Ubuntu)
+- \`dist/${RPM}\` (Fedora)
+- \`dist/SHA256SUMS.asc\` (clearsigned checksums)
+
+See [docs/install/signing-key-ceremony.md](docs/install/signing-key-ceremony.md) for key ceremony details.
+See [docs/install/migrate-from-makeinstall.md](docs/install/migrate-from-makeinstall.md) for migration from make install."
+
+if $DRY_RUN; then
+    _run "curl --fail -X POST -u ${GITEA_OWNER}:\$GITEA_TOKEN -H 'Content-Type: application/json' -d '{\"tag_name\":\"v${VERSION}\",\"name\":\"v${VERSION}\",\"body\":\"...\"}' '${GITEA_URL}/api/v1/repos/${GITEA_OWNER}/${GITEA_REPO}/releases'"
+else
+    # Create release via Gitea API (creates the tag atomically — no bare tag)
+    RELEASE_RESPONSE=$(curl --fail -s -X POST \
+        -u "${GITEA_OWNER}:${GITEA_TOKEN}" \
+        -H "Content-Type: application/json" \
+        -d "$(jq -n \
+            --arg tag "v${VERSION}" \
+            --arg name "v${VERSION}" \
+            --arg body "$_release_notes" \
+            '{tag_name: $tag, name: $name, body: $body}')" \
+        "${GITEA_URL}/api/v1/repos/${GITEA_OWNER}/${GITEA_REPO}/releases")
+
+    RELEASE_ID=$(echo "$RELEASE_RESPONSE" | jq -r '.id')
+    echo "  Release created: ${GITEA_URL}/${GITEA_OWNER}/${GITEA_REPO}/releases/tag/v${VERSION}"
+fi
+echo ""
+
+# ── Step 6: Attach artifacts to release ──────────────────────────────────
+
+echo "--- Attach artifacts to release ---"
+for artifact in "dist/${DEB}" "dist/${RPM}" "dist/SHA256SUMS.asc"; do
+    _run "curl --fail -X POST -u ${GITEA_OWNER}:\$GITEA_TOKEN -F 'attachment=@${artifact}' '${GITEA_URL}/api/v1/repos/${GITEA_OWNER}/${GITEA_REPO}/releases/${RELEASE_ID:-0}/assets'"
+done
+echo ""
+
+# ── Done ─────────────────────────────────────────────────────────────────
+
+echo "=== Release v${VERSION} complete ==="
+echo ""
+echo "Summary:"
+echo "  Packages: ${DEB}, ${RPM}"
+echo "  Checksums: dist/SHA256SUMS.asc"
+echo "  Registry: deb → bookworm, jammy, noble; rpm → ${RPM_GROUP}"
+echo "  Release: ${GITEA_URL}/${GITEA_OWNER}/${GITEA_REPO}/releases/tag/v${VERSION}"
+echo ""
+echo "Key ceremony: delete the private key after release."
+echo "  See docs/install/signing-key-ceremony.md"

scripts/release_request.py

@@ -0,0 +1,70 @@
+#!/usr/bin/env python3
+"""Describe the Gitea request that creates or resynchronizes a release."""
+from __future__ import annotations
+
+import argparse
+import json
+from pathlib import Path
+import re
+import sys
+from typing import Any
+
+
+_SEMVER = r"(?:0|[1-9]\d*)\.(?:0|[1-9]\d*)\.(?:0|[1-9]\d*)"
+
+
+class ReleaseRequestError(ValueError):
+    """A release request could not be prepared safely."""
+
+
+def build_release_request(
+    version: str, body: str, existing_release: dict[str, Any] | None
+) -> dict[str, Any]:
+    """Return the observable POST or PATCH request for a Gitea release."""
+    if not re.fullmatch(_SEMVER, version):
+        raise ReleaseRequestError(f"version is not bare semver: {version!r}")
+    if existing_release is None:
+        return {
+            "method": "POST",
+            "path": "/releases",
+            "payload": {"tag_name": f"v{version}", "name": f"v{version}", "body": body},
+        }
+
+    release_id = existing_release.get("id")
+    if not isinstance(release_id, int):
+        raise ReleaseRequestError("existing release does not contain an integer id")
+    return {
+        "method": "PATCH",
+        "path": f"/releases/{release_id}",
+        "payload": {"body": body},
+    }
+
+
+def _read_json(path: Path) -> dict[str, Any]:
+    try:
+        value = json.loads(path.read_text(encoding="utf-8"))
+    except (OSError, json.JSONDecodeError) as error:
+        raise ReleaseRequestError(f"cannot read existing release {path}: {error}") from error
+    if not isinstance(value, dict):
+        raise ReleaseRequestError("existing release must be a JSON object")
+    return value
+
+
+def main(argv: list[str] | None = None) -> int:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--version", required=True)
+    parser.add_argument("--body-file", type=Path, required=True)
+    parser.add_argument("--existing-release", type=Path)
+    args = parser.parse_args(argv)
+    try:
+        body = args.body_file.read_text(encoding="utf-8")
+        existing = _read_json(args.existing_release) if args.existing_release else None
+        print(json.dumps(build_release_request(args.version, body, existing)))
+    except (OSError, ReleaseRequestError) as error:
+        print(f"::error::{error}", file=sys.stderr)
+        return 1
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

src/fenris/__init__.py

@@ -1,2 +1,2 @@
 """Fenris: NVMe wear monitor with persistent TUI."""
-__version__ = "0.3.0"
+__version__ = "0.3.1"

src/fenris/collect.py

@@ -0,0 +1,137 @@
+#!/usr/bin/env python3
+"""fenris-collect: device interrogation and store writes.
+
+This is the root oneshot unit's ExecStart. It reads the device selector
+from /etc/fenris/fenris.conf, interrogates the drive via smartctl and sysfs,
+and writes the sample to the observation store.
+
+Spec: §8.4, §8.5
+
+When run as a script, uses the fenris package from the installed wheel.
+"""
+import json
+import subprocess
+import sys
+from datetime import datetime, timezone
+from pathlib import Path
+
+# Package dependencies are vendored independently of the host Python minor
+# version.  Keep the venv fallback for the legacy development install.
+VENV_DIR = Path("/opt/fenris")
+if VENV_DIR.exists():
+    vendor_dir = VENV_DIR / "vendor"
+    if vendor_dir.is_dir():
+        sys.path.insert(0, str(vendor_dir))
+    else:
+        site_packages = next((VENV_DIR / "lib").glob("python*/site-packages"), None)
+        if site_packages:
+            sys.path.insert(0, str(site_packages))
+
+from fenris.store import init_store, get_store_path
+from fenris.collector import run_collection
+
+
+CONFIG_PATH = Path("/etc/fenris/fenris.conf")
+
+
+def load_config() -> dict:
+    """Load configuration from /etc/fenris/fenris.conf.
+
+    The file holds exactly one key: the device selector.
+    Spec §8.3: re-read every run; no reload path.
+    """
+    if not CONFIG_PATH.exists():
+        raise RuntimeError(f"Configuration file not found: {CONFIG_PATH}")
+
+    config = {}
+    try:
+        with open(CONFIG_PATH, "r") as f:
+            for line in f:
+                line = line.strip()
+                if not line or line.startswith("#"):
+                    continue
+                if "=" in line:
+                    key, value = line.split("=", 1)
+                    config[key.strip()] = value.strip()
+    except Exception as e:
+        raise RuntimeError(f"Failed to read configuration: {e}")
+
+    if "device" not in config:
+        raise RuntimeError("Configuration error: missing 'device' key")
+
+    return config
+
+
+def interrogate_drive(device: str) -> dict:
+    """Interrogate the drive via smartctl.
+
+    Returns the smartctl JSON output.
+    Raises RuntimeError on failure.
+    """
+    result = subprocess.run(
+        ["smartctl", "-a", "-j", device],
+        capture_output=True,
+        text=True,
+    )
+
+    if result.returncode != 0:
+        raise RuntimeError(
+            f"smartctl failed for {device}: {result.stderr}"
+        )
+
+    try:
+        return json.loads(result.stdout)
+    except json.JSONDecodeError as e:
+        raise RuntimeError(f"Failed to parse smartctl output: {e}")
+
+
+def find_nvme_sysfs() -> Path | None:
+    """Find the NVMe controller sysfs path."""
+    nvme_ctrl = Path("/sys/class/nvme")
+    if not nvme_ctrl.exists():
+        return None
+
+    for ctrl in sorted(nvme_ctrl.iterdir()):
+        if ctrl.name.startswith("nvme"):
+            return ctrl
+    return None
+
+
+def main() -> None:
+    """Run one collection cycle."""
+    try:
+        config = load_config()
+        device = config["device"]
+
+        # Interrogate the drive
+        smartctl_data = interrogate_drive(device)
+
+        # Find sysfs path
+        sysfs_path = find_nvme_sysfs()
+        if sysfs_path is None:
+            raise RuntimeError("No NVMe controller found in sysfs")
+
+        # Inject a simple clock
+        class SimpleClock:
+            def utcnow(self):
+                return datetime.now(timezone.utc)
+
+        clock = SimpleClock()
+
+        # Run collection
+        result = run_collection(smartctl_data, sysfs_path, config, clock)
+
+        if result["ok"]:
+            print(f"Collection successful: {result['sample_count']} sample(s)")
+            sys.exit(0)
+        else:
+            print(f"Collection failed: {result['error']}", file=sys.stderr)
+            sys.exit(1)
+
+    except Exception as e:
+        print(f"Collection error: {e}", file=sys.stderr)
+        sys.exit(1)
+
+
+if __name__ == "__main__":
+    main()

src/fenris/collector.py

@@ -16,6 +16,8 @@ from pathlib import Path
 from typing import Any, Dict, Optional, Tuple
 
 from .store import init_store, get_store_path
+from .monitoring_periods import ensure_period_open
+from .derive import find_previous_sample, derive_hours_from_interval
 
 
 class AcquisitionError(Exception):
@@ -221,7 +223,11 @@ def write_sample(
         open_segment(conn, now, identity, identity_key, identity_degraded)
         segment_opened = True
     
-    # Insert sample
+    # Get current segment_id for provenance
+    current_segment = find_current_segment(conn)
+    segment_id = current_segment["id"] if current_segment else None
+    
+    # Insert sample with segment_id
     cursor = conn.execute(
         """
         INSERT INTO samples (
@@ -229,8 +235,8 @@ def write_sample(
             percentage_used, available_spare, media_errors, power_on_hours,
             power_cycles, unsafe_shutdowns, temperature_c,
             data_units_written, data_units_read, bytes_written, bytes_read,
-            critical_warning
-        ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+            critical_warning, segment_id
+        ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
         """,
         (
             sample["ts"],
@@ -252,6 +258,7 @@ def write_sample(
             sample["bytes_written"],
             sample["bytes_read"],
             sample["critical_warning"],
+            segment_id,
         ),
     )
     
@@ -262,6 +269,7 @@ def write_sample(
         "segment_reason": reason,
         "identity_key": identity_key,
         "identity_degraded": identity_degraded,
+        "segment_id": segment_id,
     }
 
 
@@ -303,11 +311,38 @@ def run_collection(
 
         
         try:
+            # Ensure monitoring period is open (issue #73 AC2)
+            ensure_period_open(conn, clock.utcnow())
+            
             # Validate invariants
             validate_sample_invariants(sample, conn)
             
-            # Write sample
-            write_sample(sample, identity, conn, clock)
+            # Write sample and get segment info
+            seg_info = write_sample(sample, identity, conn, clock)
+            
+            # Derive hour observations from interval with previous sample
+            try:
+                # Find the sample we just wrote
+                cursor = conn.execute("SELECT id FROM samples ORDER BY id DESC LIMIT 1")
+                current_id = cursor.fetchone()[0]
+                
+                prev = find_previous_sample(conn, seg_info.get("segment_id"), current_id)
+                if prev is not None:
+                    # Build current sample dict for derivation
+                    current = {
+                        "id": current_id,
+                        "ts": sample["ts"],
+                        "bytes_written": sample["bytes_written"],
+                        "bytes_read": sample["bytes_read"],
+                        "power_on_hours": sample["power_on_hours"],
+                        "temperature_c": sample["temperature_c"],
+                        "data_units_written": sample["data_units_written"],
+                        "data_units_read": sample["data_units_read"],
+                    }
+                    derive_hours_from_interval(conn, prev, current)
+            except Exception:
+                # Derivation failure must not prevent sample persistence (issue #73 AC6)
+                pass
             
             return {
                 "ok": True,

src/fenris/derive.py

@@ -0,0 +1,237 @@
+"""Interval derivation: samples → hour observations → day aggregates.
+
+After each collection run, the collector calls into this module to:
+1. Find the previous sample in the same segment
+2. Compute deltas (bytes, POH, temperature)
+3. Classify the hour(s) the interval spans
+4. Write/update hour_observations for each affected hour
+5. Update day_aggregates with unattributed cross-hour bytes
+
+Cross-hour deltas are retained once with unknown shares explicit (issue #73 AC4).
+No proportional allocation, endpoint assignment, or double counting.
+"""
+import sqlite3
+from datetime import datetime, timedelta, timezone
+from typing import Any, Dict, List, Optional, Tuple
+
+from .hour_classify import classify_hour, HourSplit
+
+
+def find_previous_sample(
+    conn: sqlite3.Connection,
+    segment_id: Optional[int],
+    current_sample_id: int,
+) -> Optional[Dict[str, Any]]:
+    """Find the most recent sample before current_sample_id in the same segment.
+
+    Returns None if no previous sample exists (first sample in segment).
+    """
+    if segment_id is not None:
+        cursor = conn.execute(
+            "SELECT id, ts, bytes_written, bytes_read, power_on_hours, "
+            "       temperature_c, data_units_written, data_units_read "
+            "FROM samples WHERE id < ? AND segment_id = ? "
+            "ORDER BY id DESC LIMIT 1",
+            (current_sample_id, segment_id),
+        )
+    else:
+        cursor = conn.execute(
+            "SELECT id, ts, bytes_written, bytes_read, power_on_hours, "
+            "       temperature_c, data_units_written, data_units_read "
+            "FROM samples WHERE id < ? "
+            "ORDER BY id DESC LIMIT 1",
+            (current_sample_id,),
+        )
+    row = cursor.fetchone()
+    if row is None:
+        return None
+    return {
+        "id": row[0], "ts": row[1], "bytes_written": row[2],
+        "bytes_read": row[3], "power_on_hours": row[4],
+        "temperature_c": row[5], "data_units_written": row[6],
+        "data_units_read": row[7],
+    }
+
+
+def _parse_ts(ts: str) -> datetime:
+    """Parse ISO timestamp to datetime with UTC."""
+    dt = datetime.fromisoformat(ts)
+    if dt.tzinfo is None:
+        dt = dt.replace(tzinfo=timezone.utc)
+    return dt
+
+
+def _hour_floor(dt: datetime) -> datetime:
+    """Floor a datetime to its UTC hour boundary."""
+    return dt.replace(minute=0, second=0, microsecond=0)
+
+
+def _hours_spanned(start: datetime, end: datetime) -> List[datetime]:
+    """Return list of UTC hour boundaries spanned by [start, end)."""
+    hours = []
+    h = _hour_floor(start)
+    while h < end:
+        hours.append(h)
+        h += timedelta(hours=1)
+    return hours
+
+
+def _compute_sampled_seconds_in_hour(
+    start: datetime, end: datetime, hour_start: datetime
+) -> int:
+    """How many seconds of the sample interval fall within this hour."""
+    hour_end = hour_start + timedelta(hours=1)
+    effective_start = max(start, hour_start)
+    effective_end = min(end, hour_end)
+    if effective_start >= effective_end:
+        return 0
+    return int((effective_end - effective_start).total_seconds())
+
+
+def derive_hours_from_interval(
+    conn: sqlite3.Connection,
+    prev_sample: Dict[str, Any],
+    next_sample: Dict[str, Any],
+) -> List[Dict[str, Any]]:
+    """Derive hour observations from a sample pair interval.
+
+    Returns list of hour observation dicts that were written/updated.
+    """
+    prev_ts = _parse_ts(prev_sample["ts"])
+    next_ts = _parse_ts(next_sample["ts"])
+
+    # Deltas
+    bw_delta = max(0, next_sample["bytes_written"] - prev_sample["bytes_written"])
+    br_delta = max(0, next_sample["bytes_read"] - prev_sample["bytes_read"])
+    poh_delta_s = max(0, (next_sample["power_on_hours"] - prev_sample["power_on_hours"])) * 3600
+
+    hours = _hours_spanned(prev_ts, next_ts)
+    total_span_s = int((next_ts - prev_ts).total_seconds())
+
+    results = []
+
+    if len(hours) == 1:
+        # Same-hour interval: fully attributed to this hour
+        hour_key = hours[0].strftime("%Y-%m-%dT%H:00:00+00:00")
+        sampled_s = total_span_s
+
+        # Classify hour
+        split = classify_hour(
+            wall_clock_seconds=3600,
+            poh_delta=poh_delta_s,
+            duw_delta=bw_delta,
+            dur_delta=br_delta,
+            sampled_seconds=sampled_s,
+        )
+
+        _upsert_hour_observation(
+            conn, hour_key, split,
+            bw_delta, br_delta,
+            prev_sample.get("temperature_c"), next_sample.get("temperature_c"),
+            2,  # 2 samples contributed (prev + next)
+        )
+        results.append({"hour": hour_key, "bytes_written": bw_delta, "attributed": True})
+
+    elif len(hours) >= 2:
+        # Cross-hour interval: split wall-clock time, bytes unattributed
+        for h in hours:
+            hour_key = h.strftime("%Y-%m-%dT%H:00:00+00:00")
+            sampled_s = _compute_sampled_seconds_in_hour(prev_ts, next_ts, h)
+
+            # For cross-hour, we classify based on time only (no byte attribution)
+            # The hour gets its time split but NOT the byte delta
+            split = classify_hour(
+                wall_clock_seconds=3600,
+                poh_delta=0,  # POH attribution unknown for cross-hour
+                duw_delta=0,  # Bytes unattributed
+                dur_delta=0,
+                sampled_seconds=sampled_s,
+            )
+
+            _upsert_hour_observation(
+                conn, hour_key, split,
+                0, 0,  # No byte attribution for cross-hour
+                None, None,
+                0,  # No sample falls IN this hour
+            )
+            results.append({"hour": hour_key, "bytes_written": 0, "attributed": False})
+
+        # Track unattributed bytes at day level
+        _add_unattributed_bytes(conn, prev_ts, next_ts, bw_delta, br_delta)
+
+    return results
+
+
+def _upsert_hour_observation(
+    conn: sqlite3.Connection,
+    hour_key: str,
+    split: HourSplit,
+    bw_delta: int,
+    br_delta: int,
+    temp_min: Optional[int],
+    temp_max: Optional[int],
+    sample_count: int,
+) -> None:
+    """Insert or update an hour observation."""
+    # Check if hour exists
+    existing = conn.execute(
+        "SELECT id, bytes_written_delta, sample_count FROM hour_observations WHERE hour = ?",
+        (hour_key,),
+    ).fetchone()
+
+    if existing is None:
+        temp_avg = ((temp_min or 0) + (temp_max or 0)) / 2 if temp_min is not None else None
+        coverage = (split.seconds_active + split.seconds_idle + split.seconds_powered_off) / 3600.0
+        conn.execute(
+            """INSERT INTO hour_observations
+               (hour, active_seconds, idle_seconds, powered_off_seconds, unknown_seconds,
+                bytes_written_delta, bytes_read_delta,
+                temperature_min, temperature_avg, temperature_max,
+                sample_count, coverage)
+               VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)""",
+            (hour_key, split.seconds_active, split.seconds_idle,
+             split.seconds_powered_off, split.seconds_unknown,
+             bw_delta, br_delta,
+             temp_min, temp_avg, temp_max,
+             sample_count, coverage),
+        )
+    else:
+        # Merge: accumulate bytes and sample count
+        new_bw = existing[1] + bw_delta
+        new_samples = existing[2] + sample_count
+        conn.execute(
+            "UPDATE hour_observations SET bytes_written_delta = ?, sample_count = ? WHERE id = ?",
+            (new_bw, new_samples, existing[0]),
+        )
+    conn.commit()
+
+
+def _add_unattributed_bytes(
+    conn: sqlite3.Connection,
+    prev_ts: datetime,
+    next_ts: datetime,
+    bw_delta: int,
+    br_delta: int,
+) -> None:
+    """Add unattributed byte deltas to day aggregates for each day touched."""
+    prev_day = prev_ts.strftime("%Y-%m-%d")
+    next_day = next_ts.strftime("%Y-%m-%d")
+
+    days = {prev_day, next_day}
+    for day in days:
+        existing = conn.execute(
+            "SELECT id FROM day_aggregates WHERE day = ?", (day,)
+        ).fetchone()
+        if existing is None:
+            conn.execute(
+                "INSERT INTO day_aggregates (day, unattributed_bytes_written, unattributed_bytes_read) "
+                "VALUES (?, ?, ?)",
+                (day, bw_delta, br_delta),
+            )
+        else:
+            conn.execute(
+                "UPDATE day_aggregates SET unattributed_bytes_written = unattributed_bytes_written + ?, "
+                "unattributed_bytes_read = unattributed_bytes_read + ? WHERE day = ?",
+                (bw_delta, br_delta, day),
+            )
+    conn.commit()

src/fenris/monitor.py

@@ -0,0 +1,283 @@
+#!/usr/bin/env python3
+"""fenris-monitor: privileged helper for toggle, collect, and baseline operations.
+
+This binary is the ONLY sanctioned control path for:
+- enable/disable (toggle) with monitoring-period bookkeeping
+- on-demand collection trigger
+- baseline set/clear persistence
+
+Polkit authorizes this binary under com.bongbetic.fenris.monitor (auth_admin).
+
+Spec: §8.4, §8.5, §8.6, §8.7
+
+When run as a script, uses the fenris package from the installed wheel.
+"""
+import argparse
+import json
+import os
+import subprocess
+import sys
+import sqlite3
+from datetime import datetime, timezone
+from pathlib import Path
+
+# Package dependencies are vendored independently of the host Python minor
+# version.  Keep the venv fallback for the legacy development install.
+VENV_DIR = Path("/opt/fenris")
+if VENV_DIR.exists():
+    vendor_dir = VENV_DIR / "vendor"
+    if vendor_dir.is_dir():
+        sys.path.insert(0, str(vendor_dir))
+    else:
+        site_packages = next((VENV_DIR / "lib").glob("python*/site-packages"), None)
+        if site_packages:
+            sys.path.insert(0, str(site_packages))
+
+from fenris.store import init_store, get_store_path
+from fenris.monitoring_periods import (
+    ensure_period_open,
+    close_period,
+    get_open_period,
+)
+
+
+DEFAULT_STORE_PATH = Path("/var/lib/fenris/observations.db")
+
+
+def is_root() -> bool:
+    """Check if running as root."""
+    return os.geteuid() == 0
+
+
+def cmd_enable(args: argparse.Namespace) -> None:
+    """Enable monitoring: enable timer + open monitoring period.
+
+    Idempotent matrix (§8.6):
+    - First-ever enable: opens a period at the enable moment
+    - Resume with open period: no-op (gap stays inside as unknown)
+    - Resume with no open period: opens a new row
+    """
+    store_path = getattr(args, 'store_path', DEFAULT_STORE_PATH)
+    conn = init_store(store_path)
+    now = datetime.now(timezone.utc)
+
+    try:
+        # Open monitoring period if none exists (§8.6)
+        open_period = get_open_period(conn)
+        if open_period is None:
+            ensure_period_open(conn, now)
+            print("Monitoring period opened at", now.isoformat())
+        else:
+            print("Monitoring period already open (id=%d)" % open_period["id"])
+
+        # Enable and start the timer
+        if args.now:
+            result = subprocess.run(
+                ["systemctl", "enable", "--now", "fenris-collect.timer"],
+                capture_output=True,
+                text=True,
+            )
+        else:
+            result = subprocess.run(
+                ["systemctl", "enable", "fenris-collect.timer"],
+                capture_output=True,
+                text=True,
+            )
+
+        if result.returncode != 0:
+            print("Error enabling timer:", result.stderr, file=sys.stderr)
+            sys.exit(1)
+
+        print("Timer enabled" + (" and started" if args.now else ""))
+    finally:
+        conn.close()
+
+
+def cmd_disable(args: argparse.Namespace) -> None:
+    """Disable monitoring: disable timer + close monitoring period.
+
+    Idempotent matrix (§8.6):
+    - Pause with open period: closes it user_disabled
+    - Pause otherwise: no-op
+    """
+    store_path = getattr(args, 'store_path', DEFAULT_STORE_PATH)
+    if not store_path.exists():
+        print("Error: Observation store not found at", store_path, file=sys.stderr)
+        sys.exit(1)
+
+    conn = init_store(store_path)
+    now = datetime.now(timezone.utc)
+
+    try:
+        # Close monitoring period if open (§8.6)
+        open_period = get_open_period(conn)
+        if open_period is not None:
+            close_period(conn, now, "user_disabled")
+            print("Monitoring period closed (id=%d)" % open_period["id"])
+        else:
+            print("No open monitoring period (no-op)")
+
+        # Disable and stop the timer
+        if args.now:
+            result = subprocess.run(
+                ["systemctl", "disable", "--now", "fenris-collect.timer"],
+                capture_output=True,
+                text=True,
+            )
+        else:
+            result = subprocess.run(
+                ["systemctl", "disable", "fenris-collect.timer"],
+                capture_output=True,
+                text=True,
+            )
+
+        if result.returncode != 0:
+            print("Error disabling timer:", result.stderr, file=sys.stderr)
+            sys.exit(1)
+
+        print("Timer disabled" + (" and stopped" if args.now else ""))
+    finally:
+        conn.close()
+
+
+def cmd_collect(args: argparse.Namespace) -> None:
+    """Trigger on-demand collection.
+
+    Starts fenris-collect.service, blocks until exit, reports outcome.
+
+    Spec §8.7: fenris sample routes through fenris-monitor → systemctl start,
+    which blocks until the oneshot exits; outcome reported synchronously.
+    """
+    result = subprocess.run(
+        ["systemctl", "start", "fenris-collect.service"],
+        capture_output=True,
+        text=True,
+    )
+
+    if result.returncode == 0:
+        print("Collection completed successfully")
+    else:
+        print("Collection failed:", result.stderr, file=sys.stderr)
+        sys.exit(1)
+
+
+def cmd_baseline_set(args: argparse.Namespace) -> None:
+    """Persist a baseline row after CLI-side validation.
+
+    Spec §8.4: fenris-monitor persists CLI-validated baseline rows.
+    Spec PR-14: baseline set persists through polkit-guarded helper.
+    """
+    store_path = getattr(args, 'store_path', DEFAULT_STORE_PATH)
+    if not store_path.exists():
+        print("Error: Observation store not found at", store_path, file=sys.stderr)
+        sys.exit(1)
+
+    conn = init_store(store_path)
+    now = datetime.now(timezone.utc)
+
+    try:
+        # Parse and validate baseline data
+        data = json.loads(args.baseline_json)
+
+        required_fields = [
+            "tbw_terabytes",
+            "source_url",
+            "document_revision",
+            "entry_date",
+            "model_string",
+            "nominal_capacity_bytes",
+        ]
+        for field in required_fields:
+            if field not in data:
+                print(f"Error: Missing required field: {field}", file=sys.stderr)
+                sys.exit(1)
+
+        # One active row replaced on edit (§6.2)
+        conn.execute("DELETE FROM endurance_baseline")
+        conn.execute(
+            """
+            INSERT INTO endurance_baseline (
+                tbw_terabytes, source_url, document_revision,
+                entry_date, model_string, nominal_capacity_bytes,
+                validated_by, verified, created_at, updated_at
+            ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+            """,
+            (
+                data["tbw_terabytes"],
+                data["source_url"],
+                data["document_revision"],
+                data["entry_date"],
+                data["model_string"],
+                data["nominal_capacity_bytes"],
+                data.get("validated_by", "user"),
+                data.get("verified", False),
+                now.isoformat(),
+                now.isoformat(),
+            ),
+        )
+        conn.commit()
+        print("Baseline persisted")
+    finally:
+        conn.close()
+
+
+def cmd_baseline_clear(args: argparse.Namespace) -> None:
+    """Clear the endurance baseline.
+
+    Spec PR-14: baseline clear persists through polkit-guarded helper.
+    """
+    store_path = getattr(args, 'store_path', DEFAULT_STORE_PATH)
+    if not store_path.exists():
+        print("Error: Observation store not found at", store_path, file=sys.stderr)
+        sys.exit(1)
+
+    conn = init_store(store_path)
+    try:
+        conn.execute("DELETE FROM endurance_baseline")
+        conn.commit()
+        print("Baseline cleared")
+    finally:
+        conn.close()
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(
+        prog="fenris-monitor",
+        description="Fenris privileged helper for toggle, collect, and baseline operations.",
+    )
+    subparsers = parser.add_subparsers(dest="command", required=True)
+
+    # enable/disable
+    enable_parser = subparsers.add_parser("enable", help="Enable monitoring")
+    enable_parser.add_argument(
+        "--now", action="store_true", help="Also start the timer immediately"
+    )
+    enable_parser.set_defaults(func=cmd_enable)
+
+    disable_parser = subparsers.add_parser("disable", help="Disable monitoring")
+    disable_parser.add_argument(
+        "--now", action="store_true", help="Also stop the timer immediately"
+    )
+    disable_parser.set_defaults(func=cmd_disable)
+
+    # collect
+    collect_parser = subparsers.add_parser("collect", help="Trigger on-demand collection")
+    collect_parser.set_defaults(func=cmd_collect)
+
+    # baseline
+    baseline_parser = subparsers.add_parser("baseline", help="Baseline operations")
+    baseline_sub = baseline_parser.add_subparsers(dest="baseline_action", required=True)
+
+    baseline_set = baseline_sub.add_parser("set", help="Persist baseline")
+    baseline_set.add_argument("baseline_json", help="Baseline JSON data")
+    baseline_set.set_defaults(func=cmd_baseline_set)
+
+    baseline_clear = baseline_sub.add_parser("clear", help="Clear baseline")
+    baseline_clear.set_defaults(func=cmd_baseline_clear)
+
+    args = parser.parse_args()
+    args.func(args)
+
+
+if __name__ == "__main__":
+    main()

src/fenris/preferences.py

@@ -0,0 +1,98 @@
+"""User-scoped TUI preferences (issue #80).
+
+Persists theme preset and reduced-motion choice per unprivileged user.
+Preferences live at XDG_CONFIG_HOME/fenris/preferences.json and must not
+affect collection, projection, history evidence, helper state, package
+config, or CLI status.
+
+Safe failures: invalid/unreadable/unwritable data never crashes the
+dashboard, corrupts previous preferences, or affects monitoring.
+Failures are understandable rather than silently implying persistence
+succeeded.
+
+Criteria: TPH-10, AC80-2, AC80-3.
+"""
+import json
+import os
+from pathlib import Path
+from typing import Any, Dict
+
+
+PREFERENCE_FILE_NAME = "preferences.json"
+VALID_THEMES = {"amber", "nord", "high_contrast"}
+DEFAULT_THEME = "amber"
+DEFAULT_REDUCED_MOTION = False
+
+
+def get_preference_path() -> Path:
+    """Return the user-scoped preference file path.
+
+    Uses XDG_CONFIG_HOME/fenris/preferences.json.
+    Falls back to ~/.config/fenris/preferences.json if unset.
+    """
+    xdg = os.environ.get("XDG_CONFIG_HOME")
+    if xdg:
+        base = Path(xdg)
+    else:
+        base = Path.home() / ".config"
+    return base / "fenris" / PREFERENCE_FILE_NAME
+
+
+def load_preferences() -> Dict[str, Any]:
+    """Load user preferences with safe defaults.
+
+    Returns a dict with keys:
+      theme: str (one of VALID_THEMES)
+      reduced_motion: bool
+
+    If the file is missing, corrupt, unreadable, or contains invalid
+    values, returns safe defaults (Amber theme, normal motion).
+    """
+    path = get_preference_path()
+
+    try:
+        text = path.read_text()
+    except (OSError, FileNotFoundError):
+        return _defaults()
+
+    try:
+        data = json.loads(text)
+    except (json.JSONDecodeError, ValueError):
+        return _defaults()
+
+    if not isinstance(data, dict):
+        return _defaults()
+
+    theme = data.get("theme", DEFAULT_THEME)
+    if theme not in VALID_THEMES:
+        theme = DEFAULT_THEME
+
+    reduced_motion = data.get("reduced_motion", DEFAULT_REDUCED_MOTION)
+    if not isinstance(reduced_motion, bool):
+        reduced_motion = DEFAULT_REDUCED_MOTION
+
+    return {"theme": theme, "reduced_motion": reduced_motion}
+
+
+def save_preferences(theme: str = DEFAULT_THEME,
+                     reduced_motion: bool = DEFAULT_REDUCED_MOTION) -> None:
+    """Save user preferences.
+
+    Creates the config directory if needed.  If the write fails
+    (read-only filesystem, permissions), the failure is swallowed —
+    the TUI continues with whatever was loaded, and the user sees
+    no crash or error.
+    """
+    path = get_preference_path()
+
+    try:
+        path.parent.mkdir(parents=True, exist_ok=True)
+        payload = json.dumps({"theme": theme, "reduced_motion": reduced_motion}, indent=2)
+        path.write_text(payload + "\n")
+    except (OSError, PermissionError):
+        # Best-effort persistence — failure must not crash the TUI
+        pass
+
+
+def _defaults() -> Dict[str, Any]:
+    return {"theme": DEFAULT_THEME, "reduced_motion": DEFAULT_REDUCED_MOTION}

src/fenris/projection.py

@@ -73,6 +73,7 @@ class ScenarioRange:
     rates: Dict[int, float]
     min_days: int
     max_days: int
+    horizon_reasons: Dict[int, str] = field(default_factory=dict)
 
 
 @dataclass(frozen=True)
@@ -91,6 +92,7 @@ class ProjectionResult:
     staleness_fact: Optional[str]
     degraded_identity_fact: Optional[str]
     zero_rate_fact: Optional[str]
+    qualifying_days_progress: Optional[str] = None  # Issue #77: honest qualifying-day progress
 
 
 # ---------------------------------------------------------------------------
@@ -234,20 +236,39 @@ def _compute_regime_rate(days, conn, regime_start_day, clock_now):
     return regime_bytes / regime_wc, regime_bytes, regime_wc
 
 
-def _compute_horizon_rate(days, conn, horizon_days, clock_now):
-    cutoff = (clock_now - timedelta(days=horizon_days)).strftime("%Y-%m-%d")
+def _compute_horizon_rate(days, conn, horizon_days, evidence_endpoint):
+    """Compute horizon rate anchored at the latest evidence endpoint T.
+
+    Uses exact trailing horizon_days × 86400 seconds from T, not clock_now.
+    Reader refresh alone never moves T or dilutes rates.
+
+    Returns (rate, reason) where reason is None on success or a string
+    describing why the rate is unavailable.
+    """
+    if not days:
+        return None, "no observation history"
+
+    # T is the latest evidence endpoint — the end of the last day aggregate
+    T = datetime.fromisoformat(days[-1]["day"] + "T23:59:59+00:00")
+
+    # Exact trailing start: T minus horizon_days × 86400 seconds
+    h_start = T - timedelta(days=horizon_days)
+    cutoff = h_start.strftime("%Y-%m-%d")
+
     # History must span the full horizon — no placeholders
-    if not days or days[0]["day"] > cutoff:
-        return None
+    if days[0]["day"] > cutoff:
+        return None, "%d-day window starts before earliest data" % horizon_days
+
     h_bytes = sum(d["bytes_written"] for d in days if d["day"] >= cutoff)
     covered = sum(1 for d in days if d["day"] >= cutoff)
     if covered == 0:
-        return None
-    h_start = datetime.fromisoformat(cutoff + "T00:00:00+00:00")
-    h_wc = _wall_clock_in_range(conn, h_start, clock_now)
+        return None, "%d-day window has no data" % horizon_days
+
+    h_wc = _wall_clock_in_range(conn, h_start, T)
     if h_wc <= 0:
-        return None
-    return h_bytes / h_wc
+        return None, "%d-day window has no monitored wall-clock time" % horizon_days
+
+    return h_bytes / h_wc, None
 
 
 # ---------------------------------------------------------------------------
@@ -508,17 +529,35 @@ def compute_projection(conn, clock_now):
 
     scenario = None
     horizon_rates = {}
+    horizon_reasons = {}
     for h in HORIZON_DAYS:
-        hr = _compute_horizon_rate(all_days, conn, h, clock_now)
+        hr, reason = _compute_horizon_rate(all_days, conn, h, clock_now)
         if hr is not None:
             horizon_rates[h] = hr
+        else:
+            horizon_reasons[h] = reason
     if horizon_rates:
-        scenario = ScenarioRange(rates=horizon_rates, min_days=min(horizon_rates), max_days=max(horizon_rates))
+        scenario = ScenarioRange(
+            rates=horizon_rates,
+            min_days=min(horizon_rates),
+            max_days=max(horizon_rates),
+            horizon_reasons=horizon_reasons,
+        )
 
     total_days_count = len(segment_days)
     days_below_coverage = sum(1 for d in segment_days
                              if d["coverage"] < WARMING_COVERAGE_FLOOR or d["sample_count"] == 0)
     qualifying = total_days_count - days_below_coverage
+    
+    # Issue #77: Show honest qualifying-day progress
+    if total_days_count >= WARMING_MIN_DAYS:
+        # After warm-up, show qualifying day details for transparency
+        qualifying_days_progress = "%d of %d qualifying days" % (qualifying, total_days_count)
+        if days_below_coverage > 0:
+            qualifying_days_progress += " (%d below coverage)" % days_below_coverage
+    else:
+        qualifying_days_progress = None
+    
     if total_days_count < WARMING_MIN_DAYS or days_below_coverage > WARMING_MAX_LOW_COVERAGE:
         warming_fact = "warming up: %d of %d qualifying days" % (qualifying, WARMING_MIN_DAYS)
         facts.append(warming_fact)
@@ -577,4 +616,5 @@ def compute_projection(conn, clock_now):
         staleness_fact=staleness_fact,
         degraded_identity_fact=degraded_identity_fact,
         zero_rate_fact=zero_rate_fact,
+        qualifying_days_progress=qualifying_days_progress,
     )

src/fenris/pruning.py

@@ -2,6 +2,7 @@
 
 Raw samples are pruned opportunistically to 14 days.
 Hour observations and day aggregates are retained indefinitely.
+Boundary anchors required for successor evidence are retained.
 """
 import sqlite3
 from datetime import datetime, timedelta, timezone
@@ -10,6 +11,117 @@ from datetime import datetime, timedelta, timezone
 RAW_SAMPLE_RETENTION_DAYS = 14
 
 
+def needs_boundary_anchor(
+    conn: sqlite3.Connection,
+    sample_ts: str,
+    now: datetime,
+) -> bool:
+    """Check if a sample is needed as a boundary anchor for derivation.
+
+    A sample is a boundary anchor if:
+    1. It's older than retention_days (strictly before cutoff)
+    2. It has a next sample that forms an interval spanning the retention boundary
+    3. The interval hasn't been derived yet
+
+    The interval spans the boundary if:
+    - The sample is before the cutoff, AND
+    - The next sample is strictly after the cutoff (or within retention)
+    """
+    from .derive import _parse_ts
+    
+    sample_dt = _parse_ts(sample_ts)
+    retention_cutoff = now - timedelta(days=RAW_SAMPLE_RETENTION_DAYS)
+    
+    # If sample is within retention (strictly after cutoff), not an anchor
+    if sample_dt > retention_cutoff:
+        return False
+    
+    # Check if this sample has a next sample
+    cursor = conn.execute(
+        """SELECT ts, segment_id FROM samples WHERE ts > ? ORDER BY ts LIMIT 1""",
+        (sample_ts,),
+    )
+    next_row = cursor.fetchone()
+    
+    if next_row is None:
+        # No next sample - this is the last sample
+        # It's not needed for derivation (no interval to derive)
+        return False
+    
+    next_ts_str = next_row[0]
+    next_segment_id = next_row[1]
+    next_dt = _parse_ts(next_ts_str)
+    
+    # Check if the next sample is strictly after the cutoff (i.e., interval spans boundary)
+    if next_dt > retention_cutoff:
+        # The interval spans the retention boundary
+        # Check if the interval needs derivation
+        # Get current sample's segment_id
+        cursor = conn.execute(
+            "SELECT segment_id FROM samples WHERE ts = ?",
+            (sample_ts,),
+        )
+        current_segment_row = cursor.fetchone()
+        current_segment_id = current_segment_row[0] if current_segment_row else None
+        
+        # If different segments, no interval to derive
+        if current_segment_id != next_segment_id:
+            return False
+        
+        # Check if the interval [sample_ts, next_ts] needs derivation
+        # It needs derivation if any hour in the span lacks an observation
+        current_hour = sample_dt.replace(minute=0, second=0, microsecond=0)
+        end_hour = next_dt.replace(minute=0, second=0, microsecond=0)
+        
+        while current_hour <= end_hour:
+            cursor = conn.execute(
+                "SELECT id FROM hour_observations WHERE hour = ?",
+                (current_hour.isoformat(),),
+            )
+            if cursor.fetchone() is None:
+                # This hour lacks an observation - interval needs derivation
+                return True
+            current_hour += timedelta(hours=1)
+        
+        # All hours in the span have observations - interval is derived
+        return False
+    else:
+        # The interval doesn't span the boundary (both samples are old)
+        # Check if the interval needs derivation
+        # Get current sample's segment_id
+        cursor = conn.execute(
+            "SELECT segment_id FROM samples WHERE ts = ?",
+            (sample_ts,),
+        )
+        current_segment_row = cursor.fetchone()
+        current_segment_id = current_segment_row[0] if current_segment_row else None
+        
+        # If different segments, no interval to derive
+        if current_segment_id != next_segment_id:
+            return False
+        
+        # Check if the interval [sample_ts, next_ts] needs derivation
+        current_hour = sample_dt.replace(minute=0, second=0, microsecond=0)
+        end_hour = next_dt.replace(minute=0, second=0, microsecond=0)
+        
+        while current_hour <= end_hour:
+            cursor = conn.execute(
+                "SELECT id FROM hour_observations WHERE hour = ?",
+                (current_hour.isoformat(),),
+            )
+            if cursor.fetchone() is None:
+                # This hour lacks an observation - interval needs derivation
+                # But only keep if the interval is significant (spans multiple hours)
+                # or if the next sample is the last sample before a gap
+                gap = (next_dt - sample_dt).total_seconds()
+                if gap > 24 * 3600:  # Significant gap (> 24 hours)
+                    return True
+            current_hour += timedelta(hours=1)
+        
+        # All hours in the span have observations or gap is not significant
+        return False
+
+
 def prune_old_samples(
     conn: sqlite3.Connection,
     now: datetime,
@@ -17,6 +129,8 @@ def prune_old_samples(
 ) -> int:
     """Remove raw samples older than retention_days.
 
+    Retains boundary anchors required for successor evidence.
+
     Args:
         conn: Connection to the observation store.
         now: Current UTC time.
@@ -26,6 +140,23 @@ def prune_old_samples(
         Number of samples removed.
     """
     cutoff = (now - timedelta(days=retention_days)).isoformat()
-    cursor = conn.execute("DELETE FROM samples WHERE ts < ?", (cutoff,))
+    
+    # Get all samples older than cutoff
+    cursor = conn.execute(
+        "SELECT id, ts FROM samples WHERE ts < ? ORDER BY ts",
+        (cutoff,),
+    )
+    old_samples = cursor.fetchall()
+    
+    removed = 0
+    for sample_id, sample_ts in old_samples:
+        # Check if this sample is a boundary anchor
+        if needs_boundary_anchor(conn, sample_ts, now):
+            continue  # Skip - it's a boundary anchor
+        
+        # Remove the sample
+        conn.execute("DELETE FROM samples WHERE id = ?", (sample_id,))
+        removed += 1
+    
     conn.commit()
-    return cursor.rowcount
+    return removed

src/fenris/repair.py

@@ -0,0 +1,448 @@
+"""Historical repair and retention (issue #74).
+
+Re-derives hour observations and day aggregates from surviving raw samples,
+with idempotent and interruption-safe guarantees. Preserves import markers,
+boundary anchors, and valid historical summaries.
+
+Contracts:
+- Idempotent: running repair multiple times produces no duplicates
+- Interruption-safe: partial repair preserves prior valid history
+- Preserves existing valid data: never overwrites valid derived data
+- Surfaces failures explicitly for retry
+"""
+import logging
+import sqlite3
+from dataclasses import dataclass, field
+from datetime import datetime, timedelta, timezone
+from typing import Optional, List, Tuple
+
+from .derive import find_previous_sample, derive_hours_from_interval, _parse_ts
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class RepairResult:
+    """Result of a repair operation."""
+    ok: bool
+    hours_created: int = 0
+    hours_updated: int = 0
+    days_created: int = 0
+    days_updated: int = 0
+    intervals_derived: int = 0
+    boundary_anchors_retained: int = 0
+    legacy_summaries_preserved: int = 0
+    error: Optional[str] = None
+
+
+@dataclass
+class RepairStatus:
+    """Current repair status for read-only views."""
+    last_repair: Optional[str] = None  # ISO timestamp of last successful repair
+    repair_in_progress: bool = False
+    hours_derived: int = 0
+    days_derived: int = 0
+
+
+def _ensure_repair_metadata(conn: sqlite3.Connection) -> None:
+    """Ensure metadata table exists for tracking repair state."""
+    conn.execute("""
+        CREATE TABLE IF NOT EXISTS store_metadata (
+            key TEXT PRIMARY KEY,
+            value TEXT NOT NULL
+        )
+    """)
+
+
+def is_repair_in_progress(conn: sqlite3.Connection) -> bool:
+    """Check if a repair operation is currently in progress."""
+    _ensure_repair_metadata(conn)
+    cursor = conn.execute(
+        "SELECT value FROM store_metadata WHERE key = 'repair_in_progress'"
+    )
+    row = cursor.fetchone()
+    return row is not None and row[0] == "true"
+
+
+def _set_repair_in_progress(conn: sqlite3.Connection, in_progress: bool) -> None:
+    """Mark repair as in progress or complete."""
+    _ensure_repair_metadata(conn)
+    conn.execute(
+        "INSERT OR REPLACE INTO store_metadata (key, value) VALUES (?, ?)",
+        ("repair_in_progress", "true" if in_progress else "false"),
+    )
+    conn.commit()
+
+
+def _update_repair_status(conn: sqlite3.Connection, result: RepairResult) -> None:
+    """Update repair status after successful completion."""
+    _ensure_repair_metadata(conn)
+    now = datetime.now(timezone.utc).isoformat()
+    
+    # Update last repair timestamp
+    conn.execute(
+        "INSERT OR REPLACE INTO store_metadata (key, value) VALUES (?, ?)",
+        ("last_repair", now),
+    )
+    
+    # Update derived counts
+    cursor = conn.execute("SELECT COUNT(*) FROM hour_observations")
+    hours = cursor.fetchone()[0]
+    
+    cursor = conn.execute("SELECT COUNT(*) FROM day_aggregates")
+    days = cursor.fetchone()[0]
+    
+    conn.execute(
+        "INSERT OR REPLACE INTO store_metadata (key, value) VALUES (?, ?)",
+        ("hours_derived", str(hours)),
+    )
+    conn.execute(
+        "INSERT OR REPLACE INTO store_metadata (key, value) VALUES (?, ?)",
+        ("days_derived", str(days)),
+    )
+    
+    conn.commit()
+
+
+def get_repair_status(conn: sqlite3.Connection) -> RepairStatus:
+    """Get current repair status for read-only views."""
+    _ensure_repair_metadata(conn)
+    
+    last_repair = None
+    cursor = conn.execute(
+        "SELECT value FROM store_metadata WHERE key = 'last_repair'"
+    )
+    row = cursor.fetchone()
+    if row:
+        last_repair = row[0]
+    
+    in_progress = is_repair_in_progress(conn)
+    
+    hours_derived = 0
+    cursor = conn.execute(
+        "SELECT value FROM store_metadata WHERE key = 'hours_derived'"
+    )
+    row = cursor.fetchone()
+    if row:
+        hours_derived = int(row[0])
+    
+    days_derived = 0
+    cursor = conn.execute(
+        "SELECT value FROM store_metadata WHERE key = 'days_derived'"
+    )
+    row = cursor.fetchone()
+    if row:
+        days_derived = int(row[0])
+    
+    return RepairStatus(
+        last_repair=last_repair,
+        repair_in_progress=in_progress,
+        hours_derived=hours_derived,
+        days_derived=days_derived,
+    )
+
+
+def needs_boundary_anchor(
+    conn: sqlite3.Connection,
+    sample_ts: str,
+    now: datetime,
+) -> bool:
+    """Check if a sample is needed as a boundary anchor for derivation.
+
+    A sample is a boundary anchor if:
+    1. It's older than retention_days
+    2. It has no derived hour observation for its hour
+    3. It's the last sample before a gap that needs derivation (gap > 24 hours)
+    """
+    from .pruning import RAW_SAMPLE_RETENTION_DAYS
+    
+    sample_dt = _parse_ts(sample_ts)
+    retention_cutoff = now - timedelta(days=RAW_SAMPLE_RETENTION_DAYS)
+    
+    # If sample is within retention, not an anchor (will be kept anyway)
+    if sample_dt >= retention_cutoff:
+        return False
+    
+    # Check if this sample's hour already has a derived observation
+    hour_start = sample_dt.replace(minute=0, second=0, microsecond=0).isoformat()
+    hour_end = (sample_dt + timedelta(hours=1)).replace(minute=0, second=0, microsecond=0).isoformat()
+    
+    cursor = conn.execute(
+        """SELECT COUNT(*) FROM hour_observations 
+           WHERE hour >= ? AND hour < ?""",
+        (hour_start, hour_end),
+    )
+    
+    # If there's an hour observation in this sample's hour, it's been derived
+    if cursor.fetchone()[0] > 0:
+        return False
+    
+    # Check if this sample is the last sample before a gap
+    # (i.e., the next sample is significantly later)
+    cursor = conn.execute(
+        """SELECT ts FROM samples WHERE ts > ? ORDER BY ts LIMIT 1""",
+        (sample_ts,),
+    )
+    next_row = cursor.fetchone()
+    
+    if next_row is None:
+        # No next sample - this is the last sample, might be needed
+        # But if it's old and fully derived, it's not needed
+        return False
+    
+    next_ts = _parse_ts(next_row[0])
+    gap = (next_ts - sample_dt).total_seconds()
+    
+    # If gap > 24 hours, this sample is a boundary anchor
+    # (needed to derive the interval spanning the gap)
+    return gap > 24 * 3600
+
+
+def _get_unlinked_intervals(conn: sqlite3.Connection) -> List[Tuple[dict, dict]]:
+    """Find sample pairs that form intervals but have no hour observations."""
+    cursor = conn.execute(
+        """SELECT id, ts, bytes_written, bytes_read, power_on_hours,
+                  temperature_c, data_units_written, data_units_read, segment_id
+           FROM samples ORDER BY ts"""
+    )
+    
+    all_samples = []
+    for row in cursor.fetchall():
+        all_samples.append({
+            "id": row[0], "ts": row[1], "bytes_written": row[2],
+            "bytes_read": row[3], "power_on_hours": row[4],
+            "temperature_c": row[5], "data_units_written": row[6],
+            "data_units_read": row[7], "segment_id": row[8],
+        })
+    
+    intervals = []
+    for i in range(len(all_samples) - 1):
+        prev = all_samples[i]
+        next_s = all_samples[i + 1]
+        
+        # Skip if different segments
+        if prev["segment_id"] != next_s["segment_id"]:
+            continue
+        
+        # Check if the interval spans hours that need derivation
+        prev_dt = _parse_ts(prev["ts"])
+        next_dt = _parse_ts(next_s["ts"])
+        
+        # Check if any hour in the span lacks an observation
+        current = prev_dt.replace(minute=0, second=0, microsecond=0)
+        end = next_dt.replace(minute=0, second=0, microsecond=0)
+        
+        needs_derivation = False
+        while current <= end:
+            cursor2 = conn.execute(
+                "SELECT id FROM hour_observations WHERE hour = ?",
+                (current.isoformat(),),
+            )
+            if cursor2.fetchone() is None:
+                needs_derivation = True
+                break
+            current += timedelta(hours=1)
+        
+        if needs_derivation:
+            intervals.append((prev, next_s))
+    
+    return intervals
+
+
+def _derive_day_aggregate_from_hours(
+    conn: sqlite3.Connection,
+    day: str,
+) -> Optional[dict]:
+    """Derive a day aggregate from its hour observations."""
+    cursor = conn.execute(
+        """SELECT SUM(active_seconds), SUM(idle_seconds), 
+                  SUM(powered_off_seconds), SUM(unknown_seconds),
+                  SUM(bytes_written_delta), SUM(bytes_read_delta),
+                  SUM(sample_count)
+           FROM hour_observations WHERE hour LIKE ?""",
+        (day + "T%",),
+    )
+    
+    row = cursor.fetchone()
+    if row is None or row[0] is None:
+        return None
+    
+    return {
+        "day": day,
+        "active_seconds": row[0] or 0,
+        "idle_seconds": row[1] or 0,
+        "powered_off_seconds": row[2] or 0,
+        "unknown_seconds": row[3] or 0,
+        "bytes_written_delta": row[4] or 0,
+        "bytes_read_delta": row[5] or 0,
+        "sample_count": row[6] or 0,
+    }
+
+
+def _upsert_day_aggregate(conn: sqlite3.Connection, day_data: dict) -> bool:
+    """Insert or update a day aggregate. Returns True if created."""
+    existing = conn.execute(
+        "SELECT id FROM day_aggregates WHERE day = ?",
+        (day_data["day"],),
+    ).fetchone()
+    
+    if existing is None:
+        # Calculate coverage
+        total_seconds = (day_data["active_seconds"] + day_data["idle_seconds"] + 
+                        day_data["powered_off_seconds"] + day_data["unknown_seconds"])
+        coverage = (day_data["active_seconds"] + day_data["idle_seconds"] + 
+                   day_data["powered_off_seconds"]) / total_seconds if total_seconds > 0 else 0.0
+        
+        conn.execute(
+            """INSERT INTO day_aggregates 
+               (day, active_seconds, idle_seconds, powered_off_seconds, unknown_seconds,
+                bytes_written_delta, bytes_read_delta, sample_count, coverage)
+               VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)""",
+            (day_data["day"], day_data["active_seconds"], day_data["idle_seconds"],
+             day_data["powered_off_seconds"], day_data["unknown_seconds"],
+             day_data["bytes_written_delta"], day_data["bytes_read_delta"],
+             day_data["sample_count"], coverage),
+        )
+        return True
+    else:
+        # Update existing (but only if new data is more complete)
+        # This implements the "don't overwrite valid older history" rule
+        cursor = conn.execute(
+            """SELECT bytes_written_delta, sample_count 
+               FROM day_aggregates WHERE day = ?""",
+            (day_data["day"],),
+        )
+        existing_data = cursor.fetchone()
+        
+        # Only update if new data has more samples or more bytes
+        if (day_data["sample_count"] > existing_data[1] or
+            day_data["bytes_written_delta"] > existing_data[0]):
+            
+            total_seconds = (day_data["active_seconds"] + day_data["idle_seconds"] + 
+                            day_data["powered_off_seconds"] + day_data["unknown_seconds"])
+            coverage = (day_data["active_seconds"] + day_data["idle_seconds"] + 
+                       day_data["powered_off_seconds"]) / total_seconds if total_seconds > 0 else 0.0
+            
+            conn.execute(
+                """UPDATE day_aggregates SET 
+                   active_seconds = ?, idle_seconds = ?, powered_off_seconds = ?,
+                   unknown_seconds = ?, bytes_written_delta = ?, bytes_read_delta = ?,
+                   sample_count = ?, coverage = ?
+                   WHERE day = ?""",
+                (day_data["active_seconds"], day_data["idle_seconds"],
+                 day_data["powered_off_seconds"], day_data["unknown_seconds"],
+                 day_data["bytes_written_delta"], day_data["bytes_read_delta"],
+                 day_data["sample_count"], coverage, day_data["day"]),
+            )
+            return False  # Updated, not created
+        else:
+            return False  # No update needed
+
+
+def repair_derivation(
+    conn: sqlite3.Connection,
+    clock=None,
+) -> RepairResult:
+    """Repair hour observations and day aggregates from surviving samples.
+
+    This is the main entry point for historical repair. It:
+    1. Finds sample pairs that need interval derivation
+    2. Derives hour observations from those intervals
+    3. Updates day aggregates from the hour observations
+    4. Preserves existing valid data
+    5. Is idempotent and interruption-safe
+
+    Args:
+        conn: Connection to the observation store
+        clock: Injected clock (for testing)
+
+    Returns:
+        RepairResult with operation details
+    """
+    if clock is None:
+        clock = datetime.now(timezone.utc)
+    elif hasattr(clock, 'utcnow'):
+        clock = clock.utcnow()
+    
+    # Check if repair is already in progress
+    if is_repair_in_progress(conn):
+        return RepairResult(
+            ok=False,
+            error="Repair already in progress",
+        )
+    
+    # Mark repair as in progress
+    _set_repair_in_progress(conn, True)
+    
+    result = RepairResult(ok=True)
+    
+    try:
+        # Begin transaction
+        conn.execute("BEGIN IMMEDIATE")
+        
+        # 1. Find and derive intervals from sample pairs
+        intervals = _get_unlinked_intervals(conn)
+        
+        for prev, next_s in intervals:
+            try:
+                derived_hours = derive_hours_from_interval(conn, prev, next_s)
+                result.intervals_derived += 1
+                result.hours_created += len([h for h in derived_hours if h.get("attributed", True)])
+            except Exception as e:
+                logger.warning("Failed to derive interval %s -> %s: %s",
+                             prev["ts"], next_s["ts"], e)
+                # Continue with other intervals (resilient)
+        
+        # 2. Update day aggregates from hour observations
+        cursor = conn.execute(
+            "SELECT DISTINCT substr(hour, 1, 10) as day FROM hour_observations ORDER BY day"
+        )
+        days = [row[0] for row in cursor.fetchall()]
+        
+        for day in days:
+            day_data = _derive_day_aggregate_from_hours(conn, day)
+            if day_data is not None:
+                created = _upsert_day_aggregate(conn, day_data)
+                if created:
+                    result.days_created += 1
+                else:
+                    result.days_updated += 1
+        
+        # 3. Count boundary anchors retained
+        now = clock if isinstance(clock, datetime) else datetime.now(timezone.utc)
+        cursor = conn.execute("SELECT ts FROM samples ORDER BY ts")
+        anchor_count = 0
+        for row in cursor.fetchall():
+            if needs_boundary_anchor(conn, row[0], now):
+                anchor_count += 1
+        result.boundary_anchors_retained = anchor_count
+        
+        # 4. Count preserved legacy summaries
+        # Legacy summaries are day aggregates without corresponding hour observations
+        cursor = conn.execute(
+            """SELECT COUNT(*) FROM day_aggregates d 
+               WHERE NOT EXISTS (
+                   SELECT 1 FROM hour_observations h 
+                   WHERE h.hour LIKE d.day || 'T%'
+               )"""
+        )
+        result.legacy_summaries_preserved = cursor.fetchone()[0]
+        
+        # Commit transaction
+        conn.commit()
+        
+        # Update repair status
+        _update_repair_status(conn, result)
+        
+    except Exception as e:
+        conn.rollback()
+        logger.error("Repair failed: %s", e)
+        return RepairResult(
+            ok=False,
+            error=str(e),
+        )
+    finally:
+        # Mark repair as complete
+        _set_repair_in_progress(conn, False)
+    
+    return result

src/fenris/status.py

@@ -88,7 +88,13 @@ def open_store_readonly(store_path: Path) -> sqlite3.Connection:
 
     Raises StoreFault if unreadable, NewerSchema if user_version > SCHEMA_VERSION.
     """
-    if not store_path.exists():
+    try:
+        exists = store_path.exists()
+    except OSError as e:
+        # A non-group user stat()ing a 2750 store directory gets
+        # PermissionError before any StoreFault can be raised (issue #54).
+        raise StoreFault("observation store not readable: %s" % e)
+    if not exists:
         raise StoreFault("observation store not found at %s" % store_path)
 
     try:
@@ -325,7 +331,8 @@ def check_retired_flag(flag: str) -> Optional[str]:
 def _format_projection(proj, freshness: str, service: Dict[str, Any],
                        drive_facts: List[str], config_error: Optional[str],
                        store_fault: Optional[str], newer_schema: Optional[str],
-                       journal_hint: Optional[str]) -> str:
+                       journal_hint: Optional[str],
+                       sample_count: int = 0, day_count: int = 0) -> str:
     """Format the complete status output."""
     lines = []
 
@@ -355,6 +362,16 @@ def _format_projection(proj, freshness: str, service: Dict[str, Any],
         _append_service_facts(lines, service)
         return "\n".join(lines)
 
+    # --- Single sample: awaiting another sample (issue #73 AC3) ---
+    # Only show awaiting state when there are no day aggregates (e.g., legacy import
+    # or hand-crafted stores can have 1 sample but sufficient day data for projection)
+    if sample_count <= 1 and day_count == 0:
+        lines.append("awaiting another sample")
+        lines.append("")
+        lines.append("Collecting usage data — the first projection requires at least two samples.")
+        _append_service_facts(lines, service)
+        return "\n".join(lines)
+
     # --- Projection headline ---
     headline = _format_headline(proj)
     lines.append(headline)
@@ -362,21 +379,30 @@ def _format_projection(proj, freshness: str, service: Dict[str, Any],
 
     # --- Confidence state + contributing facts (§6.7, §6.11) ---
     state_label = proj.confidence_state.value
-    if proj.contributing_facts:
-        facts_str = " · ".join(proj.contributing_facts)
+    facts_list = list(proj.contributing_facts) if proj.contributing_facts else []
+    
+    # Issue #77: Show qualifying day progress for honesty
+    if proj.qualifying_days_progress:
+        facts_list.insert(0, proj.qualifying_days_progress)
+    
+    if facts_list:
+        facts_str = " · ".join(facts_list)
         lines.append("%s evidence · %s" % (state_label, facts_str))
     else:
         lines.append("%s evidence" % state_label)
     lines.append("")
 
-    # --- Scenario range (§6.5) ---
-    if proj.scenario_range and proj.scenario_range.rates:
+    # --- Scenario range (§6.5) with horizon reasons ---
+    if proj.scenario_range:
         parts = []
         for horizon in sorted(proj.scenario_range.rates.keys()):
             rate_gb_day = proj.scenario_range.rates[horizon] * 86400 / 1e9
             parts.append("%dd: %.2f GB/day" % (horizon, rate_gb_day))
-        lines.append("scenario range: %s" % " · ".join(parts))
-        lines.append("")
+        for horizon, reason in sorted(proj.scenario_range.horizon_reasons.items()):
+            parts.append("%dd: %s" % (horizon, reason))
+        if parts:
+            lines.append("scenario range: %s" % " · ".join(parts))
+            lines.append("")
 
     # --- PU context line (§6.1) ---
     lines.append(proj.pu_context_line)
@@ -443,7 +469,7 @@ def _format_headline(proj) -> str:
 
 
 def _append_service_facts(lines: List[str], service: Dict[str, Any]) -> None:
-    """Append the four separate service facts (§7.3, LC-9)."""
+    """Append service facts and dashboard-clarity monitoring state."""
     boot = "enabled" if service.get("boot_enabled") else "disabled"
     activity = "active" if service.get("timer_active") else "inactive"
 
@@ -467,6 +493,55 @@ def _append_service_facts(lines: List[str], service: Dict[str, Any]) -> None:
 
     lines.append("boot: %s · timer: %s · last collect: %s%s · freshness: %s%s"
                  % (boot, activity, collect, collect_age, freshness_str, freshness_age))
+    lines.append("CONTINUITY: %s" % monitoring_continuity(service))
+    if service.get("deliberately_paused"):
+        lines.extend(deliberate_pause_lines())
+
+
+# ---------------------------------------------------------------------------
+# Dashboard clarity parity wording (DC-2, DC-3)
+# ---------------------------------------------------------------------------
+
+_CONTINUITY_ACTIVE = "monitoring: active in background · persists across reboots"
+_CONTINUITY_DISABLED = "monitoring: does not start on next boot"
+_PAUSED_TITLE = "monitoring: paused — deliberate disable"
+_PAUSED_CONSEQUENCE = (
+    "paused time is excluded from your usage habit · resume: fenris monitor resume"
+)
+
+
+def monitoring_continuity(service: Dict[str, Any]) -> str:
+    """Return the boot-persistence wording, independent of timer runtime."""
+    return _CONTINUITY_ACTIVE if service.get("boot_enabled") else _CONTINUITY_DISABLED
+
+
+def deliberate_pause_lines() -> List[str]:
+    """Return the exact CLI/TUI presentation for a sanctioned pause."""
+    return [_PAUSED_TITLE, _PAUSED_CONSEQUENCE]
+
+
+def is_deliberately_paused(conn: sqlite3.Connection, service: Dict[str, Any]) -> bool:
+    """Whether the latest closed period was ended by Fenris's own pause path.
+
+    Raw systemd operations have no `user_disabled` row, so they must never be
+    presented as a Deliberate disable. A live enabled timer also wins over a
+    stale period marker, keeping the presentation consistent with service facts.
+    """
+    if service.get("boot_enabled") or service.get("timer_active"):
+        return False
+
+    open_period = conn.execute(
+        "SELECT 1 FROM monitoring_periods WHERE ended_at IS NULL LIMIT 1"
+    ).fetchone()
+    if open_period is not None:
+        return False
+
+    row = conn.execute(
+        "SELECT end_cause FROM monitoring_periods "
+        "WHERE ended_at IS NOT NULL "
+        "ORDER BY ended_at DESC, id DESC LIMIT 1"
+    ).fetchone()
+    return row is not None and row[0] == "user_disabled"
 
 
 def format_disclosures() -> str:
@@ -543,6 +618,17 @@ def get_status(store_path: Optional[Path] = None, clock_now: Optional[datetime]
 
     freshness = grade_freshness(newest_ts, clock_now)
 
+    # --- Sample count for single-sample state (issue #73 AC3) ---
+    sample_count = 0
+    day_count = 0
+    try:
+        cursor = conn.execute("SELECT COUNT(*) FROM samples")
+        sample_count = cursor.fetchone()[0]
+        cursor = conn.execute("SELECT COUNT(*) FROM day_aggregates")
+        day_count = cursor.fetchone()[0]
+    except sqlite3.Error:
+        pass
+
     # Freshness age for the service fact
     freshness_age_s = None
     if newest_ts:
@@ -556,6 +642,10 @@ def get_status(store_path: Optional[Path] = None, clock_now: Optional[datetime]
 
     service["freshness"] = freshness
     service["freshness_age_s"] = freshness_age_s
+    try:
+        service["deliberately_paused"] = is_deliberately_paused(conn, service)
+    except sqlite3.Error:
+        service["deliberately_paused"] = False
 
     # --- Drive anomalies (§9.7, FL-7) ---
     drive_facts = []
@@ -578,7 +668,7 @@ def get_status(store_path: Optional[Path] = None, clock_now: Optional[datetime]
     # --- Compose output ---
     result = _format_projection(
         proj, freshness, service, drive_facts, config_error,
-        None, None, journal_hint,
+        None, None, journal_hint, sample_count, day_count,
     )
 
     conn.close()
@@ -597,3 +687,79 @@ def render_status(store_path: Optional[Path] = None, clock_now: Optional[datetim
         parts.append("")
         parts.append(format_disclosures())
     return "\n\n".join(parts)
+
+
+
+# ---------------------------------------------------------------------------
+# Shared status composition integration (issue #78)
+# ---------------------------------------------------------------------------
+
+def get_status_composition(
+    store_path: Optional[Path] = None,
+    clock_now: Optional[datetime] = None,
+    query_services: bool = True,
+    collecting: bool = False,
+    reduced_motion: bool = False,
+) -> 'StatusComposition':
+    """Get the shared status composition consumed by both TUI and CLI.
+
+    This is the new entry point that centralizes the status lattice.
+    """
+    # Lazy import to avoid circular dependency
+    from .status_composition import (
+        StatusComposition,
+        compose_status,
+    )
+
+    if clock_now is None:
+        clock_now = datetime.now(timezone.utc)
+
+    # --- Configuration (§8.3) ---
+    # Config errors are surfaced through the store fault mechanism
+
+    # --- Service state ---
+    service = {}
+    if query_services:
+        try:
+            service = query_service_state()
+        except Exception:
+            service = None
+
+    # --- Store open ---
+    store_fault = None
+    newer_schema = None
+    conn = None
+
+    if store_path is None:
+        store_path = Path("/var/lib/fenris/observations.db")
+
+    try:
+        conn = open_store_readonly(store_path)
+    except StoreFault as e:
+        store_fault = str(e)
+    except NewerSchema as e:
+        newer_schema = str(e)
+
+    # --- Use shared composition ---
+    if conn is not None:
+        try:
+            comp = compose_status(
+                conn, service, clock_now,
+                store_fault=store_fault,
+                newer_schema=newer_schema,
+                collecting=collecting,
+                reduced_motion=reduced_motion,
+            )
+        finally:
+            conn.close()
+    else:
+        # Store fault or newer schema - compose without store data
+        comp = compose_status(
+            None, service, clock_now,
+            store_fault=store_fault,
+            newer_schema=newer_schema,
+            collecting=collecting,
+            reduced_motion=reduced_motion,
+        )
+
+    return comp

src/fenris/status_composition.py

@@ -0,0 +1,526 @@
+"""Shared status composition for TUI and CLI (issue #78, TPH-2/3).
+
+Centralizes the monitoring status lattice consumed by both surfaces.
+States: Monitoring, Collecting, Paused, Waiting, Interrupted, Error, Stale, Unknown.
+
+Precedence: Error > Interrupted > Paused > Stale > Waiting > Monitoring > Unknown.
+Collecting overlays every base except store fault.
+
+Freshness grading uses shared constants from status.py.
+"""
+import enum
+import sqlite3
+from dataclasses import dataclass, field
+from datetime import datetime, timedelta, timezone
+from typing import Any, Dict, List, Optional
+
+# Status poll interval (AC78-6): lightweight 5s systemctl show poll
+STATUS_POLL_INTERVAL_S = 5
+
+from .status import (
+    FRESH_THRESHOLD_S,
+    STALENESS_THRESHOLD_S,
+    grade_freshness,
+    freshness_age_human,
+    is_deliberately_paused,
+    monitoring_continuity,
+    deliberate_pause_lines,
+    open_store_readonly,
+    StoreFault,
+    NewerSchema,
+)
+
+
+# ---------------------------------------------------------------------------
+# Status state enum with glyph and label
+# ---------------------------------------------------------------------------
+
+class StatusState(enum.Enum):
+    """The eight monitoring status states (TPH-2)."""
+
+    MONITORING = "monitoring"
+    COLLECTING = "collecting"
+    PAUSED = "paused"
+    WAITING = "waiting"
+    INTERRUPTED = "interrupted"
+    ERROR = "error"
+    STALE = "stale"
+    UNKNOWN = "unknown"
+
+    @property
+    def glyph(self) -> str:
+        _glyphs = {
+            "monitoring": "●",
+            "collecting": "◐",
+            "paused": "‖",
+            "waiting": "○",
+            "interrupted": "⊘",
+            "error": "✖",
+            "stale": "◌",
+            "unknown": "?",
+        }
+        return _glyphs[self.value]
+
+    @property
+    def label(self) -> str:
+        return self.value.capitalize()
+
+
+# Precedence order: higher index = higher precedence
+_PRECEDENCE = [
+    StatusState.UNKNOWN,
+    StatusState.MONITORING,
+    StatusState.WAITING,
+    StatusState.STALE,
+    StatusState.PAUSED,
+    StatusState.INTERRUPTED,
+    StatusState.ERROR,
+]
+
+_PRECEDENCE_RANK = {s: i for i, s in enumerate(_PRECEDENCE)}
+
+
+# ---------------------------------------------------------------------------
+# Status composition result
+# ---------------------------------------------------------------------------
+
+@dataclass
+class StatusComposition:
+    """The composed status result shared between TUI and CLI."""
+
+    state: StatusState
+    glyph: str
+    label: str
+    explanation: str
+
+    # Separate facts (never folded into the status word)
+    freshness: str = "unknown"
+    freshness_age_s: Optional[int] = None
+    last_collect_ok: Optional[bool] = None
+    last_collect_age_s: Optional[int] = None
+    last_collect_reason: Optional[str] = None
+    boot_enabled: Optional[bool] = None
+    timer_active: Optional[bool] = None
+    deliberately_paused: bool = False
+    pause_age_s: Optional[int] = None
+    external_stop_reason: Optional[str] = None
+
+    # Store fault / newer schema (suppress store-dependent views)
+    store_fault: Optional[str] = None
+    newer_schema: Optional[str] = None
+
+    # Collecting overlay
+    overlay_base: Optional[StatusState] = None
+
+    # Sample counts for waiting explanations
+    sample_count: int = 0
+    day_count: int = 0
+
+    # Whether the status dot should blink (only Monitoring)
+    should_blink: bool = False
+
+    # Continuity line
+    continuity: str = ""
+
+    # Paused lines (for TUI banner)
+    paused_lines: List[str] = field(default_factory=list)
+
+
+# ---------------------------------------------------------------------------
+# Status composition logic
+# ---------------------------------------------------------------------------
+
+def _determine_base_state(
+    freshness: str,
+    sample_count: int,
+    day_count: int,
+    boot_enabled: Optional[bool],
+    timer_active: Optional[bool],
+    last_collect_ok: Optional[bool],
+    deliberately_paused: bool,
+    external_stop_reason: Optional[str],
+    service_available: bool,
+) -> StatusState:
+    """Determine the base status state from facts.
+
+    Precedence: Error > Interrupted > Paused > Stale > Waiting > Monitoring > Unknown.
+    """
+    # Unknown: service query failed and no store-derived fact places us higher
+    if not service_available:
+        return StatusState.UNKNOWN
+
+    # Error: last collect failed
+    if last_collect_ok is False:
+        return StatusState.ERROR
+
+    # Interrupted: external stop (not user_disabled)
+    if external_stop_reason is not None:
+        return StatusState.INTERRUPTED
+
+    # Paused: deliberate disable
+    if deliberately_paused:
+        return StatusState.PAUSED
+
+    # Stale: timer active, no failure, but data ≥ 48h old
+    if freshness == "stale" and timer_active and last_collect_ok is not False:
+        return StatusState.STALE
+
+    # Waiting: empty store, single sample, or fresh data but not yet enough evidence
+    if sample_count == 0:
+        return StatusState.WAITING
+    if sample_count <= 1 and day_count == 0:
+        return StatusState.WAITING
+
+    # Monitoring: everything is fine
+    return StatusState.MONITORING
+
+
+def _determine_explanation(
+    state: StatusState,
+    freshness: str,
+    freshness_age_s: Optional[int],
+    last_collect_ok: Optional[bool],
+    last_collect_reason: Optional[str],
+    sample_count: int,
+    deliberately_paused: bool,
+    store_fault: Optional[str],
+    newer_schema: Optional[str],
+) -> str:
+    """Determine the explanation line for the status state."""
+    if store_fault:
+        return "observation store unreadable — see journal"
+
+    if newer_schema:
+        return "observation store written by a newer Fenris — upgrade Fenris"
+
+    if state == StatusState.ERROR:
+        parts = []
+        if last_collect_ok is False:
+            parts.append("last run failed")
+            if last_collect_reason:
+                parts.append("(%s)" % last_collect_reason)
+        if freshness_age_s is not None and freshness != "empty":
+            parts.append("· last good sample %s ago" % freshness_age_human(freshness_age_s))
+        return " ".join(parts) if parts else "last run failed"
+
+    if state == StatusState.INTERRUPTED:
+        return "collection stopped outside Fenris — monitoring period still open"
+
+    if state == StatusState.PAUSED:
+        return "monitoring paused — paused time excluded from your usage habit"
+
+    if state == StatusState.STALE:
+        if freshness_age_s is not None:
+            return "last sample %s ago" % freshness_age_human(freshness_age_s)
+        return "data is stale"
+
+    if state == StatusState.WAITING:
+        if sample_count == 0:
+            return "awaiting first sample"
+        if sample_count <= 1:
+            return "awaiting another sample"
+        return "waiting for data"
+
+    if state == StatusState.MONITORING:
+        if freshness_age_s is not None:
+            return "last sample %s ago" % freshness_age_human(freshness_age_s)
+        return "monitoring active"
+
+    if state == StatusState.UNKNOWN:
+        return "service state unavailable"
+
+    return ""
+
+
+def _determine_collecting_overlay(
+    base_state: StatusState,
+    last_collect_ok: Optional[bool],
+    deliberately_paused: bool,
+    store_fault: Optional[str],
+    newer_schema: Optional[str],
+) -> str:
+    """Determine the explanation line when Collecting overlays a base state."""
+    if store_fault or newer_schema:
+        return "run in flight — store fault"
+
+    if base_state == StatusState.PAUSED:
+        return "run in flight — paused"
+    if base_state == StatusState.INTERRUPTED:
+        return "run in flight — interrupted"
+    if base_state == StatusState.ERROR:
+        return "run in flight — retry"
+    if base_state == StatusState.STALE:
+        return "run in flight — stale data"
+    if base_state == StatusState.WAITING:
+        return "run in flight"
+    return "run in flight"
+
+
+def compose_status(
+    conn: sqlite3.Connection,
+    service: Optional[Dict[str, Any]],
+    clock_now: datetime,
+    store_fault: Optional[str] = None,
+    newer_schema: Optional[str] = None,
+    collecting: bool = False,
+    reduced_motion: bool = False,
+) -> StatusComposition:
+    """Compose the shared status from service state and store data.
+
+    This is the single entry point consumed by both TUI and CLI.
+    """
+    service_available = service is not None and len(service) > 0
+
+    # --- Separate facts from service ---
+    boot_enabled = service.get("boot_enabled") if service_available else None
+    timer_active = service.get("timer_active") if service_available else None
+    last_collect_ok = service.get("last_collect_ok") if service_available else None
+    last_collect_age_s = service.get("last_collect_age_s") if service_available else None
+    last_collect_reason = service.get("last_collect_reason") if service_available else None
+
+    # --- Freshness from store ---
+    freshness = "unknown"
+    freshness_age_s = None
+    sample_count = 0
+    day_count = 0
+    deliberately_paused = False
+
+    if store_fault is None and newer_schema is None:
+        try:
+            cursor = conn.execute("SELECT ts FROM samples ORDER BY id DESC LIMIT 1")
+            row = cursor.fetchone()
+            newest_ts = row[0] if row else None
+            freshness = grade_freshness(newest_ts, clock_now)
+            if newest_ts:
+                try:
+                    ts = datetime.fromisoformat(newest_ts)
+                    if ts.tzinfo is None:
+                        ts = ts.replace(tzinfo=timezone.utc)
+                    else:
+                        ts = ts.astimezone(timezone.utc)
+                    freshness_age_s = int((clock_now - ts).total_seconds())
+                except (ValueError, TypeError):
+                    pass
+        except sqlite3.Error:
+            freshness = "unknown"
+
+        try:
+            cursor = conn.execute("SELECT COUNT(*) FROM samples")
+            sample_count = cursor.fetchone()[0]
+            cursor = conn.execute("SELECT COUNT(*) FROM day_aggregates")
+            day_count = cursor.fetchone()[0]
+        except sqlite3.Error:
+            pass
+
+        try:
+            svc_for_pause = service if service_available else {}
+            deliberately_paused = is_deliberately_paused(conn, svc_for_pause)
+        except sqlite3.Error:
+            deliberately_paused = False
+
+    # --- External stop detection ---
+    # External stop = timer inactive + boot disabled + NOT deliberately paused
+    # + period still open (the timer was stopped but Fenris didn't close the period)
+    external_stop_reason = None
+    if not deliberately_paused and timer_active is False and boot_enabled is False:
+        # Check if there's an open monitoring period (external stop left it open)
+        try:
+            open_period = conn.execute(
+                "SELECT 1 FROM monitoring_periods WHERE ended_at IS NULL LIMIT 1"
+            ).fetchone()
+            if open_period is not None:
+                external_stop_reason = "external_stop"
+        except sqlite3.Error:
+            pass
+
+    # --- Determine base state ---
+    # Store fault and newer schema always override to ERROR
+    if store_fault is not None or newer_schema is not None:
+        base_state = StatusState.ERROR
+    else:
+        base_state = _determine_base_state(
+            freshness=freshness,
+            sample_count=sample_count,
+            day_count=day_count,
+            boot_enabled=boot_enabled,
+            timer_active=timer_active,
+            last_collect_ok=last_collect_ok,
+            deliberately_paused=deliberately_paused,
+            external_stop_reason=external_stop_reason,
+            service_available=service_available,
+        )
+
+    # --- Apply Collecting overlay ---
+    state = base_state
+    overlay_base = None
+    explanation = ""
+
+    if collecting and store_fault is None and newer_schema is None:
+        # Collecting overlays every base except store fault
+        overlay_base = base_state
+        state = StatusState.COLLECTING
+        explanation = _determine_collecting_overlay(
+            base_state, last_collect_ok, deliberately_paused,
+            store_fault, newer_schema,
+        )
+    else:
+        explanation = _determine_explanation(
+            base_state, freshness, freshness_age_s,
+            last_collect_ok, last_collect_reason,
+            sample_count, deliberately_paused,
+            store_fault, newer_schema,
+        )
+
+    # --- Continuity line ---
+    continuity = ""
+    if service_available:
+        continuity = monitoring_continuity(service)
+
+    # --- Paused lines ---
+    paused_lines = []
+    if deliberately_paused:
+        paused_lines = deliberate_pause_lines()
+
+    # Determine blink flag: only Monitoring dot blinks, never text or other states
+    should_blink = (state == StatusState.MONITORING and not reduced_motion)
+
+    return StatusComposition(
+        state=state,
+        glyph=state.glyph,
+        label=state.label,
+        explanation=explanation,
+        should_blink=should_blink,
+        freshness=freshness,
+        freshness_age_s=freshness_age_s,
+        last_collect_ok=last_collect_ok,
+        last_collect_age_s=last_collect_age_s,
+        last_collect_reason=last_collect_reason,
+        boot_enabled=boot_enabled,
+        timer_active=timer_active,
+        deliberately_paused=deliberately_paused,
+        external_stop_reason=external_stop_reason,
+        store_fault=store_fault,
+        newer_schema=newer_schema,
+        overlay_base=overlay_base,
+        sample_count=sample_count,
+        day_count=day_count,
+        continuity=continuity,
+        paused_lines=paused_lines,
+    )
+
+
+# ---------------------------------------------------------------------------
+# CLI rendering (static, no styling)
+# ---------------------------------------------------------------------------
+
+def render_status_cli(comp: StatusComposition) -> str:
+    """Render the status composition as static CLI text."""
+    lines = []
+
+    # Status line
+    lines.append("%s %s" % (comp.glyph, comp.label))
+
+    # Explanation
+    if comp.explanation:
+        lines.append(comp.explanation)
+
+    lines.append("")
+
+    # Separate facts
+    facts = []
+    if comp.freshness != "unknown":
+        facts.append("freshness: %s" % comp.freshness)
+    if comp.freshness_age_s is not None:
+        facts[-1] += " (%s)" % freshness_age_human(comp.freshness_age_s) if facts else "freshness: %s" % freshness_age_human(comp.freshness_age_s)
+    if comp.last_collect_ok is True:
+        facts.append("last collect: ok")
+    elif comp.last_collect_ok is False:
+        collect_str = "last collect: FAILED"
+        if comp.last_collect_reason:
+            collect_str += " (%s)" % comp.last_collect_reason
+        facts.append(collect_str)
+    else:
+        facts.append("last collect: unknown")
+    if comp.boot_enabled is not None:
+        facts.append("boot: %s" % ("enabled" if comp.boot_enabled else "disabled"))
+    if comp.timer_active is not None:
+        facts.append("timer: %s" % ("active" if comp.timer_active else "inactive"))
+
+    if facts:
+        lines.append(" · ".join(facts))
+
+    # Continuity
+    if comp.continuity:
+        lines.append("")
+        lines.append("CONTINUITY: %s" % comp.continuity)
+
+    # Deliberate pause
+    if comp.paused_lines:
+        for pl in comp.paused_lines:
+            lines.append(pl)
+
+    return "\n".join(lines)
+
+
+# ---------------------------------------------------------------------------
+# TUI rendering (with styling tokens)
+# ---------------------------------------------------------------------------
+
+def render_status_tui(comp: StatusComposition) -> str:
+    """Render the status composition as TUI text with Textual markup."""
+    lines = []
+
+    # Status line with color
+    color = {
+        StatusState.MONITORING: "green",
+        StatusState.COLLECTING: "green",
+        StatusState.PAUSED: "yellow",
+        StatusState.WAITING: "yellow",
+        StatusState.INTERRUPTED: "red",
+        StatusState.ERROR: "red",
+        StatusState.STALE: "red",
+        StatusState.UNKNOWN: "dim",
+    }[comp.state]
+
+    lines.append("[%s]%s %s[/%s]" % (color, comp.glyph, comp.label, color))
+
+    # Explanation
+    if comp.explanation:
+        lines.append(comp.explanation)
+
+    lines.append("")
+
+    # Separate facts
+    facts = []
+    if comp.freshness != "unknown":
+        facts.append("freshness: %s" % comp.freshness)
+    if comp.freshness_age_s is not None and facts:
+        facts[-1] += " (%s)" % freshness_age_human(comp.freshness_age_s)
+    if comp.last_collect_ok is True:
+        facts.append("last collect: ok")
+    elif comp.last_collect_ok is False:
+        collect_str = "last collect: FAILED"
+        if comp.last_collect_reason:
+            collect_str += " (%s)" % comp.last_collect_reason
+        facts.append(collect_str)
+    else:
+        facts.append("last collect: unknown")
+    if comp.boot_enabled is not None:
+        facts.append("boot: %s" % ("enabled" if comp.boot_enabled else "disabled"))
+    if comp.timer_active is not None:
+        facts.append("timer: %s" % ("active" if comp.timer_active else "inactive"))
+
+    if facts:
+        lines.append(" · ".join(facts))
+
+    # Continuity
+    if comp.continuity:
+        lines.append("")
+        lines.append("[bold]CONTINUITY[/bold]  %s" % comp.continuity)
+
+    # Deliberate pause
+    if comp.paused_lines:
+        for pl in comp.paused_lines:
+            lines.append(pl)
+
+    return "\n".join(lines)

src/fenris/store.py

@@ -12,12 +12,22 @@ from typing import Optional
 
 
 # Schema version - increment on each migration
-SCHEMA_VERSION = 1
+SCHEMA_VERSION = 2
+
+
+# Packaged default placement (spec §8.3). The config may override it, but a
+# fresh install that sets only the device selector must collect cleanly.
+DEFAULT_STORE_PATH = Path("/var/lib/fenris/observations.db")
 
 
 def get_store_path(config: dict) -> Path:
-    """Get the store path from config."""
-    return Path(config["store_path"])
+    """Get the store path from config.
+
+    Falls back to the packaged default when the config does not pin one,
+    so a fresh install whose config holds only the device selector works
+    instead of crashing with KeyError 'store_path' (issue #53).
+    """
+    return Path(config.get("store_path", DEFAULT_STORE_PATH))
 
 
 def init_store(store_path: Path) -> sqlite3.Connection:
@@ -31,6 +41,20 @@ def init_store(store_path: Path) -> sqlite3.Connection:
     # Enable WAL mode for concurrent reads during writes
     conn.execute("PRAGMA journal_mode=WAL")
 
+    # Group members (fenris group) read the live store read-only, but SQLite
+    # in WAL mode needs write access to the db and its -wal/-shm sidecars even
+    # for readers. Best effort: root-created stores stay group-accessible
+    # without relying on the creating process's umask (issue #54).
+    import os as _os
+    for sidecar in (store_path,
+                    store_path.with_name(store_path.name + "-wal"),
+                    store_path.with_name(store_path.name + "-shm")):
+        try:
+            mode = _os.stat(sidecar).st_mode & 0o777
+            _os.chmod(sidecar, mode | 0o060)
+        except OSError:
+            pass
+
     # Check if this is a new database
     cursor = conn.execute("PRAGMA user_version")
     current_version = cursor.fetchone()[0]
@@ -82,7 +106,8 @@ def _create_schema(conn: sqlite3.Connection):
             data_units_read INTEGER,
             bytes_written INTEGER,
             bytes_read INTEGER,
-            critical_warning INTEGER
+            critical_warning INTEGER,
+            segment_id INTEGER
         )
     """)
     
@@ -117,7 +142,9 @@ def _create_schema(conn: sqlite3.Connection):
             bytes_written_delta INTEGER DEFAULT 0,
             bytes_read_delta INTEGER DEFAULT 0,
             sample_count INTEGER DEFAULT 0,
-            coverage REAL DEFAULT 0.0
+            coverage REAL DEFAULT 0.0,
+            unattributed_bytes_written INTEGER DEFAULT 0,
+            unattributed_bytes_read INTEGER DEFAULT 0
         )
     """)
     
@@ -176,10 +203,73 @@ def _create_schema(conn: sqlite3.Connection):
 
 
 def _apply_migrations(conn: sqlite3.Connection, current_version: int):
-    """Apply forward-only migrations from current_version to SCHEMA_VERSION."""
-    # Future migrations will go here
-    # For now, just upgrade to current version
-    pass
+    """Apply forward-only migrations from current_version to SCHEMA_VERSION.
+    
+    Each migration step is a transactional block. Add new steps as sequential
+    elif branches when SCHEMA_VERSION increases.
+    
+    Spec: §3.6, §10.2
+    """
+    # Migration 1→2: add segment_id provenance to samples,
+    # unattributed byte tracking to day_aggregates (issue #73)
+    if current_version < 2:
+        # Defensive: only ALTER if table exists (handles minimal v1 stores)
+        tables = {row[0] for row in conn.execute(
+            "SELECT name FROM sqlite_master WHERE type='table'"
+        ).fetchall()}
+        if "samples" in tables:
+            # Check if column already exists (idempotent)
+            cols = {row[1] for row in conn.execute("PRAGMA table_info(samples)").fetchall()}
+            if "segment_id" not in cols:
+                conn.execute("ALTER TABLE samples ADD COLUMN segment_id INTEGER")
+        if "day_aggregates" in tables:
+            cols = {row[1] for row in conn.execute("PRAGMA table_info(day_aggregates)").fetchall()}
+            if "unattributed_bytes_written" not in cols:
+                conn.execute("ALTER TABLE day_aggregates ADD COLUMN unattributed_bytes_written INTEGER DEFAULT 0")
+            if "unattributed_bytes_read" not in cols:
+                conn.execute("ALTER TABLE day_aggregates ADD COLUMN unattributed_bytes_read INTEGER DEFAULT 0")
+        current_version = 2
+
+
+def migrate_to_latest(store_path: Path) -> int:
+    """Apply forward-only migrations to bring the store to SCHEMA_VERSION.
+    
+    Called by the upgrade target (§10.2). Returns the number of migration
+    steps applied. Raises ValueError on newer-schema store (§3.6, §9.5).
+    
+    Spec: §3.6, §10.2, §10.3
+    """
+    conn = sqlite3.connect(str(store_path))
+    conn.execute("PRAGMA journal_mode=WAL")
+    
+    cursor = conn.execute("PRAGMA user_version")
+    current_version = cursor.fetchone()[0]
+    
+    if current_version > SCHEMA_VERSION:
+        conn.close()
+        raise ValueError(
+            f"Observation store written by a newer Fenris (version {current_version}) "
+            f"— upgrade Fenris"
+        )
+    
+    if current_version == SCHEMA_VERSION:
+        conn.close()
+        return 0  # Already up to date
+    
+    # Version 0 means no schema — create fresh (issue #73)
+    if current_version == 0:
+        _create_schema(conn)
+        conn.execute(f"PRAGMA user_version={SCHEMA_VERSION}")
+        conn.commit()
+        conn.close()
+        return SCHEMA_VERSION
+    
+    steps = SCHEMA_VERSION - current_version
+    _apply_migrations(conn, current_version)
+    conn.execute(f"PRAGMA user_version={SCHEMA_VERSION}")
+    conn.commit()
+    conn.close()
+    return steps
 
 
 def is_store_faulty(store_path: Path) -> bool:

src/fenris/themes.py

@@ -0,0 +1,161 @@
+"""Fenris theme presets (issue #80).
+
+Three accessible colour presets: Amber (default), Nord, and High Contrast.
+Themes style chrome, borders, accents, muted text, and graph roles;
+status semantic colours/glyphs/text always win.
+
+Theme roles for graph rendering expose distinct colours per preset so
+the bar graph can reflect the user's visual preference without depending
+on graph-ticket completion.
+
+Criteria: TPH-10, AC80-1, AC80-5.
+"""
+from typing import Dict
+
+from textual.theme import Theme
+
+
+# ---------------------------------------------------------------------------
+# Status semantic colours — always win, never themed (AC80-1)
+# ---------------------------------------------------------------------------
+
+STATUS_COLORS = {
+    "monitoring": "green",
+    "collecting": "green",
+    "paused": "yellow",
+    "waiting": "yellow",
+    "interrupted": "red",
+    "error": "red",
+    "stale": "red",
+    "unknown": "dim",
+}
+
+
+# ---------------------------------------------------------------------------
+# Amber theme — warm golden tones (default, amber graph role)
+# ---------------------------------------------------------------------------
+
+_AMBER = Theme(
+    name="fenris-amber",
+    primary="#d4a017",       # warm amber
+    secondary="#c49b0a",     # darker amber
+    accent="#ffd54f",        # light amber highlight
+    warning="#e6a817",       # amber warning
+    error="#e74c3c",         # red error
+    success="#27ae60",       # green success
+    foreground="#e8e0d0",    # warm light
+    background="#1a1510",    # warm dark
+    surface="#241f16",       # warm surface
+    panel="#2a2318",         # warm panel
+    boost="#332a1c",         # warm boost
+    dark=True,
+    variables={
+        "graph-allocated": "#d4a017",
+        "graph-unallocated": "#8b6914",
+        "graph-gap": "#554422",
+        "graph-zero": "#665533",
+        "graph-partial": "#aa8822",
+        "graph-selection": "#ffd54f",
+        "border-default": "#554422",
+        "muted-text": "#887755",
+    },
+)
+
+# ---------------------------------------------------------------------------
+# Nord theme — cool blue-gray polar night palette
+# ---------------------------------------------------------------------------
+
+_NORD = Theme(
+    name="fenris-nord",
+    primary="#88c0d0",       # nord8 frost
+    secondary="#81a1c1",     # nord9
+    accent="#8fbcbb",        # nord7
+    warning="#ebcb8b",       # nord13
+    error="#bf616a",         # nord11
+    success="#a3be8c",       # nord14
+    foreground="#eceff4",    # nord6
+    background="#2e3440",    # nord0
+    surface="#3b4252",       # nord1
+    panel="#434c5e",         # nord2
+    boost="#4c566a",         # nord3
+    dark=True,
+    variables={
+        "graph-allocated": "#88c0d0",
+        "graph-unallocated": "#5e81ac",
+        "graph-gap": "#4c566a",
+        "graph-zero": "#616e88",
+        "graph-partial": "#81a1c1",
+        "graph-selection": "#8fbcbb",
+        "border-default": "#4c566a",
+        "muted-text": "#7b88a1",
+    },
+)
+
+# ---------------------------------------------------------------------------
+# High Contrast — maximum readability, pure black and white
+# ---------------------------------------------------------------------------
+
+_HIGH_CONTRAST = Theme(
+    name="fenris-high-contrast",
+    primary="#ffffff",       # pure white
+    secondary="#dddddd",     # light gray
+    accent="#ffff00",        # bright yellow
+    warning="#ff8800",       # bright orange
+    error="#ff0000",         # pure red
+    success="#00ff00",       # pure green
+    foreground="#ffffff",    # pure white
+    background="#000000",    # pure black
+    surface="#111111",       # near-black surface
+    panel="#1a1a1a",         # near-black panel
+    boost="#222222",         # near-black boost
+    dark=True,
+    variables={
+        "graph-allocated": "#ffffff",
+        "graph-unallocated": "#aaaaaa",
+        "graph-gap": "#555555",
+        "graph-zero": "#666666",
+        "graph-partial": "#cccccc",
+        "graph-selection": "#ffff00",
+        "border-default": "#ffffff",
+        "muted-text": "#aaaaaa",
+    },
+)
+
+
+# ---------------------------------------------------------------------------
+# Theme registry
+# ---------------------------------------------------------------------------
+
+THEMES = {
+    "amber": _AMBER,
+    "nord": _NORD,
+    "high_contrast": _HIGH_CONTRAST,
+}
+
+THEME_NAMES = set(THEMES.keys())
+
+
+def get_theme(name: str) -> Theme:
+    """Return a registered theme by preset name.
+
+    Unknown names fall back to Amber.
+    """
+    return THEMES.get(name, _AMBER)
+
+
+def get_graph_colors(theme_name: str) -> Dict[str, str]:
+    """Return the graph colour roles for a theme preset.
+
+    Returns a dict with keys: allocated, unallocated, gap, zero, partial,
+    selection.  Falls back to Amber for unknown names.
+    """
+    theme = get_theme(theme_name)
+    variables = theme.variables or {}
+    return {
+        "allocated": variables.get("graph-allocated", "#d4a017"),
+        "unallocated": variables.get("graph-unallocated", "#8b6914"),
+        "gap": variables.get("graph-gap", "#554422"),
+        "zero": variables.get("graph-zero", "#665533"),
+        "partial": variables.get("graph-partial", "#aa8822"),
+        "selection": variables.get("graph-selection", "#ffd54f"),
+    }

tests/conftest.py

@@ -0,0 +1,20 @@
+"""Shared test helpers for Fenris test suite."""
+from pathlib import Path
+
+import pytest
+
+REPO_ROOT = Path(__file__).resolve().parent.parent
+VERSION_FILE = REPO_ROOT / "pyproject.toml"
+
+
+def get_version() -> str:
+    """Extract version from pyproject.toml."""
+    for line in VERSION_FILE.read_text().splitlines():
+        if line.startswith("version"):
+            return line.split("=")[1].strip().strip('"')
+    raise RuntimeError("Could not determine version from pyproject.toml")
+
+
+def read(path: str | Path) -> str:
+    """Read a file relative to the repository root."""
+    return (REPO_ROOT / path).read_text()

tests/test_acceptance_sweep.py

@@ -0,0 +1,749 @@
+"""Cross-cutting acceptance sweep (issue #32).
+
+Systematic verification of every acceptance criterion that spans multiple
+subsystems.  Grouped by criterion ID; each test cites its clause.
+
+CI-1  Exhaustive state matrix: confidence × freshness × baseline tier
+CI-2  TUI/CLI parity: identical outcomes and wording
+CI-3  Prohibition set: automated structural checks
+CI-4  Required wording and six disclosures in both views
+"""
+import re
+import sqlite3
+from datetime import datetime, timedelta, timezone
+from pathlib import Path
+from unittest.mock import patch
+
+import pytest
+import sys
+
+sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
+
+from fenris.store import init_store, SCHEMA_VERSION
+from fenris.monitoring_periods import ensure_period_open
+from fenris.projection import (
+    compute_projection,
+    ConfidenceState,
+    BaselineTier,
+    DISCLOSURES,
+    STALENESS_HOURS,
+    WARMING_MIN_DAYS,
+    YOUNG_REGIME_DAYS,
+)
+from fenris.status import (
+    grade_freshness,
+    get_status,
+    render_status,
+    format_disclosures,
+    FRESH_THRESHOLD_S,
+    STALENESS_THRESHOLD_S,
+    CADENCE_DEFAULT_S,
+    ACCURACY_SEC,
+)
+from fenris.tui import (
+    FenrisTuiApp,
+    _format_remaining,
+)
+
+
+SRC_DIR = Path(__file__).parent.parent / "src"
+FENRIS_PKG = SRC_DIR / "fenris"
+
+
+def _clock(year=2026, month=9, day=30, hour=12):
+    return datetime(year, month, day, hour, 0, 0, tzinfo=timezone.utc)
+
+
+def _insert_baseline(conn, tbw_tb=1.0, verified=True,
+                      model="Samsung SSD 970 EVO Plus 1TB",
+                      source_url="https://example.com/spec",
+                      doc_rev="v1.0", entry_date="2026-01-01",
+                      nominal_cap=1024000000000):
+    conn.execute(
+        "INSERT INTO endurance_baseline "
+        "(tbw_terabytes, source_url, document_revision, entry_date, model_string, "
+        " nominal_capacity_bytes, validated_by, verified, created_at, updated_at) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)",
+        (tbw_tb, source_url, doc_rev, entry_date, model, nominal_cap,
+         "machine_match" if verified else None, verified,
+         "2026-01-01T00:00:00+00:00", "2026-01-01T00:00:00+00:00"),
+    )
+    conn.commit()
+
+
+def _insert_segment(conn, opened_at="2026-09-01T00:00:00+00:00",
+                     identity_key="nqn.test", degraded=False,
+                     mn="Samsung SSD 970 EVO Plus 1TB"):
+    conn.execute(
+        "INSERT INTO controller_segments "
+        "(opened_at, identity_key, identity_degraded, subnqn, sn, mn, fr, vid, ssvid, transport) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)",
+        (opened_at, identity_key, degraded, "nqn.test", "SN123", mn, "FW1",
+         "0x144d", "0x144d", "pcie"),
+    )
+    conn.commit()
+
+
+def _insert_day(conn, day, bw=1024*1024*100, coverage=0.95, samples=24):
+    conn.execute(
+        "INSERT INTO day_aggregates (day, active_seconds, idle_seconds, "
+        "powered_off_seconds, unknown_seconds, bytes_written_delta, "
+        "bytes_read_delta, sample_count, coverage) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)",
+        (day, 3600, 0, 0, 0, bw, 0, samples, coverage),
+    )
+    conn.commit()
+
+
+def _insert_sample(conn, ts, pu=5):
+    conn.execute(
+        "INSERT INTO samples (ts, device, data_units_written, data_units_read, "
+        "percentage_used, bytes_written, bytes_read, power_on_hours) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?)",
+        (ts, "/dev/nvme0n1", 1000000, 500000, pu, 512000000000, 256000000000, 8765),
+    )
+    conn.commit()
+
+
+def _open_period(conn, start="2026-09-01T00:00:00+00:00"):
+    ensure_period_open(conn, datetime.fromisoformat(start))
+
+
+def _setup_full_store(conn, *, baseline=True, segment=True, days=30,
+                       bw=1024*1024*100, coverage=0.95, samples_per_day=24,
+                       sample_ts="2026-09-30T10:00:00+00:00",
+                       period_start="2026-09-01T00:00:00+00:00",
+                       segment_opened="2026-09-01T00:00:00+00:00",
+                       baseline_kw=None, segment_kw=None):
+    if baseline:
+        _insert_baseline(conn, **(baseline_kw or {}))
+    if segment:
+        _insert_segment(conn, opened_at=segment_opened, **(segment_kw or {}))
+    _open_period(conn, start=period_start)
+    for i in range(days):
+        d = (datetime(2026, 9, 1) + timedelta(days=i)).strftime("%Y-%m-%d")
+        _insert_day(conn, d, bw=bw, coverage=coverage, samples=samples_per_day)
+    if sample_ts:
+        _insert_sample(conn, sample_ts)
+
+
+# ===================================================================
+# CI-1: Exhaustive state matrix
+# ===================================================================
+
+
+class TestCI1StateMatrix:
+    """Systematic walk of confidence x freshness x baseline tier combinations."""
+
+    def test_no_baseline_unavailable(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _insert_segment(conn)
+        _open_period(conn)
+        for i in range(30):
+            d = (datetime(2026, 9, 1) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100, coverage=0.95, samples=24)
+        _insert_sample(conn, "2026-09-30T10:00:00+00:00")
+        proj = compute_projection(conn, _clock())
+        assert proj.confidence_state == ConfidenceState.UNSUPPORTED
+        assert proj.headline_remaining_seconds is None
+        assert proj.baseline_tier == BaselineTier.NONE
+        conn.close()
+
+    def test_verified_baseline_possible_supported(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _setup_full_store(conn, baseline_kw=dict(tbw_tb=10.0, verified=True))
+        proj = compute_projection(conn, _clock())
+        assert proj.confidence_state == ConfidenceState.SUPPORTED
+        assert proj.baseline_tier == BaselineTier.VERIFIED
+        assert proj.headline_remaining_seconds is not None
+        conn.close()
+
+    def test_unverified_baseline_possible_limited(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _setup_full_store(conn, baseline_kw=dict(
+            tbw_tb=10.0, verified=False, source_url=None))
+        proj = compute_projection(conn, _clock())
+        assert proj.baseline_tier == BaselineTier.UNVERIFIED
+        assert proj.confidence_state != ConfidenceState.SUPPORTED
+        conn.close()
+
+    def test_model_mismatch_unavailable(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _setup_full_store(conn, baseline_kw=dict(model="Different Model"))
+        proj = compute_projection(conn, _clock())
+        assert proj.confidence_state == ConfidenceState.UNSUPPORTED
+        assert proj.baseline_tier == BaselineTier.NONE
+        conn.close()
+
+    def test_fresh_sample_grades_fresh(self, tmp_path):
+        now = _clock()
+        ts = (now - timedelta(seconds=FRESH_THRESHOLD_S - 10)).isoformat()
+        assert grade_freshness(ts, now) == "fresh"
+
+    def test_missed_sample_grades_missed(self, tmp_path):
+        now = _clock()
+        ts = (now - timedelta(hours=2)).isoformat()
+        assert grade_freshness(ts, now) == "missed"
+
+    def test_stale_sample_grades_stale(self, tmp_path):
+        now = _clock()
+        ts = (now - timedelta(hours=49)).isoformat()
+        assert grade_freshness(ts, now) == "stale"
+
+    def test_empty_store_grades_empty(self, tmp_path):
+        now = _clock()
+        assert grade_freshness(None, now) == "empty"
+
+    def test_unsupported_fresh(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _insert_segment(conn)
+        _open_period(conn)
+        for i in range(20):
+            d = (datetime(2026, 9, 10) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100)
+        fresh_ts = (_clock() - timedelta(seconds=60)).isoformat()
+        _insert_sample(conn, fresh_ts)
+        proj = compute_projection(conn, _clock())
+        assert proj.confidence_state == ConfidenceState.UNSUPPORTED
+        assert grade_freshness(fresh_ts, _clock()) == "fresh"
+        conn.close()
+
+    def test_limited_young_regime(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _insert_baseline(conn, tbw_tb=10.0, verified=True)
+        _insert_segment(conn)
+        _open_period(conn)
+        for i in range(5):
+            d = (datetime(2026, 9, 25) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100)
+        _insert_sample(conn, "2026-09-30T10:00:00+00:00")
+        proj = compute_projection(conn, _clock())
+        assert proj.confidence_state == ConfidenceState.LIMITED
+        conn.close()
+
+    def test_limited_warming(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _insert_baseline(conn, tbw_tb=10.0, verified=True)
+        _insert_segment(conn)
+        _open_period(conn)
+        for i in range(10):
+            d = (datetime(2026, 9, 20) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100, coverage=0.95, samples=24)
+        _insert_sample(conn, "2026-09-30T10:00:00+00:00")
+        proj = compute_projection(conn, _clock())
+        assert proj.confidence_state == ConfidenceState.LIMITED
+        assert proj.warming_fact is not None
+        conn.close()
+
+    def test_limited_stale_data(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _insert_baseline(conn, tbw_tb=10.0, verified=True)
+        _insert_segment(conn, opened_at="2026-08-01T00:00:00+00:00")
+        _open_period(conn, start="2026-08-01T00:00:00+00:00")
+        for i in range(30):
+            d = (datetime(2026, 8, 1) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100, coverage=0.95, samples=24)
+        stale_ts = (_clock() - timedelta(days=5)).isoformat()
+        _insert_sample(conn, stale_ts)
+        proj = compute_projection(conn, _clock())
+        assert proj.confidence_state == ConfidenceState.LIMITED
+        assert proj.staleness_fact is not None
+        conn.close()
+
+    def test_limited_degraded_identity(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _insert_baseline(conn, tbw_tb=10.0, verified=True)
+        _insert_segment(conn, identity_key=None, degraded=True)
+        _open_period(conn)
+        for i in range(30):
+            d = (datetime(2026, 9, 1) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100, coverage=0.95, samples=24)
+        _insert_sample(conn, "2026-09-30T10:00:00+00:00")
+        proj = compute_projection(conn, _clock())
+        assert proj.confidence_state == ConfidenceState.LIMITED
+        assert proj.degraded_identity_fact is not None
+        conn.close()
+
+    def test_unsupported_zero_rate(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _insert_baseline(conn, tbw_tb=10.0, verified=True)
+        _insert_segment(conn)
+        _open_period(conn)
+        for i in range(30):
+            d = (datetime(2026, 9, 1) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=0)
+        _insert_sample(conn, "2026-09-30T10:00:00+00:00")
+        proj = compute_projection(conn, _clock())
+        assert proj.confidence_state == ConfidenceState.UNSUPPORTED
+        assert proj.zero_rate_fact is not None
+        conn.close()
+
+    def test_headline_present_when_projection_exists(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _setup_full_store(conn, baseline_kw=dict(tbw_tb=10.0, verified=True))
+        proj = compute_projection(conn, _clock())
+        assert proj.headline_remaining_seconds is not None
+        assert proj.headline_remaining_seconds > 0
+        conn.close()
+
+    def test_headline_absent_when_unavailable(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _insert_segment(conn)
+        _open_period(conn)
+        proj = compute_projection(conn, _clock())
+        assert proj.headline_remaining_seconds is None
+        conn.close()
+
+    def test_headline_absent_when_zero_rate(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _insert_baseline(conn, tbw_tb=1.0, verified=True)
+        _insert_segment(conn)
+        _open_period(conn)
+        for i in range(20):
+            d = (datetime(2026, 9, 10) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=0)
+        proj = compute_projection(conn, _clock())
+        assert proj.headline_remaining_seconds is None
+        conn.close()
+
+    def test_facts_always_list(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _insert_segment(conn)
+        _open_period(conn)
+        proj = compute_projection(conn, _clock())
+        assert isinstance(proj.contributing_facts, list)
+        conn.close()
+
+    def test_facts_never_empty_for_unavailable(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _insert_segment(conn)
+        _open_period(conn)
+        proj = compute_projection(conn, _clock())
+        assert len(proj.contributing_facts) > 0
+        conn.close()
+
+    def test_confidence_never_percentage(self, tmp_path):
+        conn = init_store(tmp_path / "db")
+        _setup_full_store(conn, baseline_kw=dict(tbw_tb=10.0, verified=True))
+        proj = compute_projection(conn, _clock())
+        assert proj.confidence_state in (
+            ConfidenceState.UNSUPPORTED, ConfidenceState.LIMITED, ConfidenceState.SUPPORTED)
+        for f in proj.contributing_facts:
+            if re.match(r"^\\d+%$", f.strip()):
+                pytest.fail("Bare percentage in facts: %r" % f)
+        conn.close()
+
+    def test_status_renders_same_state_as_projection(self, tmp_path):
+        db = tmp_path / "observations.db"
+        conn = init_store(db)
+        _setup_full_store(conn, baseline_kw=dict(tbw_tb=10.0, verified=True))
+        conn.close()
+        now = _clock()
+        with patch("fenris.status.query_service_state", return_value={
+            "boot_enabled": True, "timer_active": True,
+            "last_collect_ok": True, "last_collect_age_s": 60,
+            "last_collect_reason": None,
+        }):
+            status = get_status(store_path=db, clock_now=now,
+                                query_services=True, query_journal=False)
+        assert "Supported" in status or "supported" in status.lower()
+        assert "remaining" in status.lower()
+
+
+# ===================================================================
+# CI-2: TUI/CLI parity
+# ===================================================================
+
+
+class TestCI2Parity:
+    """Verify TUI and CLI share the same constants, formatting, and wording."""
+
+    def test_freshness_constants_shared(self):
+        from fenris import tui as tui_mod
+        from fenris import status as status_mod
+        assert tui_mod.FRESH_THRESHOLD_S == status_mod.FRESH_THRESHOLD_S
+        assert tui_mod.STALENESS_THRESHOLD_S == status_mod.STALENESS_THRESHOLD_S
+
+    def test_grade_freshness_shared(self):
+        from fenris.tui import grade_freshness as tui_gf
+        from fenris.status import grade_freshness as status_gf
+        assert tui_gf is status_gf
+
+    def test_disclosures_shared(self):
+        from fenris.projection import DISCLOSURES as proj_disc
+        from fenris.status import format_disclosures
+        output = format_disclosures()
+        for d in proj_disc:
+            assert d in output
+
+    def test_status_four_facts_match_tui_strip(self, tmp_path):
+        db = tmp_path / "observations.db"
+        conn = init_store(db)
+        _insert_segment(conn)
+        _open_period(conn)
+        for i in range(20):
+            d = (datetime(2026, 9, 10) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100)
+        _insert_sample(conn, "2026-09-30T10:00:00+00:00")
+        conn.close()
+        now = _clock()
+        with patch("fenris.status.query_service_state", return_value={
+            "boot_enabled": True, "timer_active": True,
+            "last_collect_ok": True, "last_collect_age_s": 120,
+            "last_collect_reason": None,
+        }):
+            status = get_status(store_path=db, clock_now=now,
+                                query_services=True, query_journal=False)
+        assert "boot:" in status
+        assert "timer:" in status
+        assert "last collect:" in status
+        assert "freshness:" in status
+
+    def test_dashboard_clarity_parity_strings_have_one_status_source(self):
+        """DC-2/DC-3 wording originates in status and the TUI imports it."""
+        status_src = (FENRIS_PKG / "status.py").read_text()
+        tui_src = (FENRIS_PKG / "tui.py").read_text()
+        for wording in (
+            "monitoring: active in background · persists across reboots",
+            "monitoring: does not start on next boot",
+            "monitoring: paused — deliberate disable",
+            "paused time is excluded from your usage habit · resume: fenris monitor resume",
+        ):
+            assert status_src.count(wording) == 1
+            assert wording not in tui_src
+
+    @pytest.mark.asyncio
+    @pytest.mark.parametrize(
+        ("state", "service", "expected_lines"),
+        [
+            (
+                "active_enabled",
+                {"boot_enabled": True, "timer_active": True},
+                ["monitoring: active in background · persists across reboots"],
+            ),
+            (
+                "boot_disabled",
+                {"boot_enabled": False, "timer_active": False},
+                ["monitoring: does not start on next boot"],
+            ),
+            (
+                "deliberately_paused",
+                {"boot_enabled": False, "timer_active": False},
+                [
+                    "monitoring: does not start on next boot",
+                    "monitoring: paused — deliberate disable",
+                    "paused time is excluded from your usage habit · resume: fenris monitor resume",
+                ],
+            ),
+        ],
+    )
+    async def test_dashboard_clarity_monitoring_lines_match_both_views(
+        self, tmp_path, state, service, expected_lines
+    ):
+        """CI-2 synthetic-store sweep covers active, disabled, and paused states."""
+        db = tmp_path / (state + ".db")
+        conn = init_store(db)
+        if state == "active_enabled":
+            ensure_period_open(conn, _clock())
+        elif state == "deliberately_paused":
+            conn.execute(
+                "INSERT INTO monitoring_periods (started_at, ended_at, end_cause) "
+                "VALUES (?, ?, ?)",
+                ("2026-09-30T09:00:00+00:00", "2026-09-30T10:00:00+00:00", "user_disabled"),
+            )
+            conn.commit()
+        conn.close()
+        service_state = {
+            **service,
+            "last_collect_ok": None,
+            "last_collect_age_s": None,
+            "last_collect_reason": None,
+        }
+
+        with patch("fenris.status.query_service_state", return_value=service_state), patch(
+            "fenris.tui.query_service_state", return_value=service_state
+        ):
+            status = get_status(
+                store_path=db, clock_now=_clock(), query_services=True, query_journal=False
+            ).lower()
+            app = FenrisTuiApp(store_path=db)
+            async with app.run_test(size=(100, 40)):
+                tui_text = "\n".join(
+                    (
+                        str(app.query_one("#service-strip").render()),
+                        str(app.query_one("#paused-banner").render()),
+                    )
+                ).lower()
+
+        for expected in expected_lines:
+            assert expected in status
+            assert expected in tui_text
+        if state != "deliberately_paused":
+            assert "monitoring: paused — deliberate disable" not in status
+            assert "monitoring: paused — deliberate disable" not in tui_text
+
+    def test_pause_resume_action_names(self):
+        tui_keys = {b.key for b in FenrisTuiApp.BINDINGS}
+        assert "p" in tui_keys
+        assert "r" in tui_keys
+        assert "c" in tui_keys
+        assert "q" in tui_keys
+
+    def test_empty_store_greeting_both_views(self, tmp_path):
+        db = tmp_path / "observations.db"
+        init_store(db)
+        now = _clock()
+        with patch("fenris.status.query_service_state", return_value={
+            "boot_enabled": False, "timer_active": False,
+            "last_collect_ok": None, "last_collect_age_s": None,
+            "last_collect_reason": None,
+        }):
+            status = get_status(store_path=db, clock_now=now,
+                                query_services=True, query_journal=False)
+        assert "no observations yet" in status.lower()
+
+    def test_store_fault_phrase_both_views(self, tmp_path):
+        status_src = (FENRIS_PKG / "status.py").read_text()
+        tui_src = (FENRIS_PKG / "tui.py").read_text()
+        phrase = "observation store unreadable"
+        assert phrase in status_src
+        assert phrase.lower() in tui_src.lower()
+
+    def test_newer_schema_phrase_both_views(self):
+        status_src = (FENRIS_PKG / "status.py").read_text()
+        phrase = "observation store written by a newer Fenris"
+        assert phrase in status_src
+
+    def test_status_never_prompts(self):
+        status_src = (FENRIS_PKG / "status.py").read_text()
+        assert "input(" not in status_src
+
+
+# ===================================================================
+# CI-3: Prohibition set
+# ===================================================================
+
+
+class TestCI3ProhibitionSet:
+    """Structural codebase checks for every prohibition clause."""
+
+    def _read_all_sources(self):
+        files = {}
+        for py in FENRIS_PKG.glob("*.py"):
+            files[py.name] = py.read_text()
+        return files
+
+    def test_single_acquisition_path(self):
+        """Only fenris-collect may interrogate the device. [2.1, 8.7]
+
+        collector.py contains the acquisition functions; collect.py is the
+        fenris-collect entry point that invokes them.  No other module may
+        reference smartctl.
+        """
+        sources = self._read_all_sources()
+        allowed = {"collector.py", "collect.py"}
+        for name, text in sources.items():
+            if name in allowed:
+                continue
+            assert "smartctl" not in text, (
+                "%s must not contain smartctl" % name
+            )
+
+    def test_no_run_surface(self):
+        """No /run/fenris coordination surface. [1.2, 3]"""
+        sources = self._read_all_sources()
+        for name, text in sources.items():
+            assert "/run/fenris" not in text, (
+                "%s references /run/fenris" % name
+            )
+
+    def test_single_config_key(self):
+        """Config holds exactly one key: device. [8.3]"""
+        status_src = (FENRIS_PKG / "status.py").read_text()
+        in_read_config = False
+        config_keys = []
+        for line in status_src.split("\n"):
+            if "def read_config" in line:
+                in_read_config = True
+            elif in_read_config and line.strip().startswith("def "):
+                break
+            elif in_read_config and "key ==" in line:
+                match = re.search(r'key\s*==\s*["\']([^"\']+)["\']', line)
+                if match:
+                    config_keys.append(match.group(1))
+        assert "device" in config_keys
+        assert len(config_keys) == 1, "Found keys: %s" % config_keys
+
+    def test_no_alerting_machinery(self):
+        """No alerting, notification, or escalation. [9.6]"""
+        sources = self._read_all_sources()
+        alert_keywords = ["send_email", "smtp", "webhook", "push_notification"]
+        for name, text in sources.items():
+            for kw in alert_keywords:
+                for line in text.split("\n"):
+                    stripped = line.strip()
+                    if kw in stripped and not stripped.startswith("#"):
+                        pytest.fail(
+                            "%s contains alerting keyword '%s': %s" % (name, kw, stripped)
+                        )
+
+    def test_no_synthetic_baselines(self):
+        """No synthetic or capacity-derived baseline. [6.1]"""
+        proj_src = (FENRIS_PKG / "projection.py").read_text()
+        assert "synthetic" not in proj_src.lower()
+
+    def test_no_stored_projections(self):
+        """Projections never stored; recomputed on read. [3.7, 6.10]"""
+        store_src = (FENRIS_PKG / "store.py").read_text()
+        create_tables = re.findall(r"CREATE TABLE.*?(?=\n\n|$)", store_src, re.DOTALL)
+        table_names = []
+        for ct in create_tables:
+            m = re.search(r"IF NOT EXISTS\s+(\w+)", ct)
+            if m:
+                table_names.append(m.group(1))
+        assert "projection" not in [t.lower() for t in table_names]
+
+    def test_no_partial_newer_schema_interpretation(self):
+        """Readers refuse newer-schema stores. [3.6, 9.5]"""
+        status_src = (FENRIS_PKG / "status.py").read_text()
+        assert "NewerSchema" in status_src
+        assert "upgrade Fenris" in status_src
+
+    def test_polkit_authorizes_one_binary(self):
+        """Polkit authorizes exactly one binary: fenris-monitor. [8.5]"""
+        monitor_src = (FENRIS_PKG / "monitor.py").read_text()
+        assert "fenris-monitor" in monitor_src or "fenris_monitor" in monitor_src
+        collect_src = (FENRIS_PKG / "collect.py").read_text()
+        assert "polkit" not in collect_src.lower()
+
+    def test_no_hour_interpolation(self):
+        """No absent hour is interpolated or fabricated. [5.3]"""
+        proj_src = (FENRIS_PKG / "projection.py").read_text()
+        assert "interpolat" not in proj_src.lower()
+        assert "fabricat" not in proj_src.lower()
+
+    def test_fenris_sh_not_shipped(self):
+        """fenris.sh is not shipped. [8.8]"""
+        repo_root = Path(__file__).parent.parent
+        assert not (repo_root / "fenris.sh").exists()
+
+
+# ===================================================================
+# CI-4: Required wording and six disclosures
+# ===================================================================
+
+
+class TestCI4WordingAndDisclosures:
+    """Verify exact fixed phrases and disclosures in both views."""
+
+    def test_exactly_six_disclosures(self):
+        assert len(DISCLOSURES) == 6
+
+    def test_disclosure_1_endurance_not_failure(self):
+        assert "endurance projection" in DISCLOSURES[0].lower()
+        assert "hardware-failure" in DISCLOSURES[0].lower() or "failure date" in DISCLOSURES[0].lower()
+
+    def test_disclosure_2_vendor_specific(self):
+        assert "vendor-specific" in DISCLOSURES[1]
+        assert "255 is saturated" in DISCLOSURES[1]
+
+    def test_disclosure_3_warranty_not_failure(self):
+        assert "warranty" in DISCLOSURES[2].lower() or "endurance threshold" in DISCLOSURES[2].lower()
+        assert "failure threshold" in DISCLOSURES[2].lower()
+
+    def test_disclosure_4_duw_rounding(self):
+        assert "DUW" in DISCLOSURES[3]
+        assert "upward-rounded" in DISCLOSURES[3]
+        assert "NAND" in DISCLOSURES[3]
+
+    def test_disclosure_5_quality_depends(self):
+        assert "baseline provenance" in DISCLOSURES[4]
+        assert "future workload" in DISCLOSURES[4]
+
+    def test_disclosure_6_gaps_and_disabled(self):
+        assert "Gaps" in DISCLOSURES[5]
+        assert "deliberately disabled" in DISCLOSURES[5]
+
+    def test_disclosures_render_in_status(self):
+        output = format_disclosures()
+        assert output.startswith("Disclosures")
+        for i in range(1, 7):
+            assert "%d." % i in output
+        for d in DISCLOSURES:
+            assert d in output
+
+    def test_disclosures_render_in_tui(self):
+        tui_src = (FENRIS_PKG / "tui.py").read_text()
+        assert "format_disclosures" in tui_src
+
+    def test_zero_rate_phrase(self):
+        phrase = "no finite projection from this history"
+        proj_src = (FENRIS_PKG / "projection.py").read_text()
+        assert phrase in proj_src
+        status_src = (FENRIS_PKG / "status.py").read_text()
+        assert phrase in status_src
+
+    def test_unavailable_no_baseline_phrase(self):
+        phrase = "no applicable endurance baseline"
+        proj_src = (FENRIS_PKG / "projection.py").read_text()
+        assert phrase in proj_src
+
+    def test_store_fault_phrase(self):
+        phrase = "observation store unreadable"
+        status_src = (FENRIS_PKG / "status.py").read_text()
+        assert phrase in status_src
+        tui_src = (FENRIS_PKG / "tui.py").read_text()
+        assert phrase.lower() in tui_src.lower()
+
+    def test_newer_schema_phrase(self):
+        phrase = "observation store written by a newer Fenris"
+        status_src = (FENRIS_PKG / "status.py").read_text()
+        assert phrase in status_src
+
+    def test_no_observations_phrase(self):
+        phrase = "no observations yet"
+        status_src = (FENRIS_PKG / "status.py").read_text()
+        assert phrase in status_src
+        tui_src = (FENRIS_PKG / "tui.py").read_text()
+        assert phrase in tui_src.lower()
+
+    def test_config_error_phrase(self):
+        phrase = "configuration error:"
+        status_src = (FENRIS_PKG / "status.py").read_text()
+        assert phrase in status_src
+
+    def test_degraded_identity_phrase(self):
+        phrase = "controller identity unavailable"
+        proj_src = (FENRIS_PKG / "projection.py").read_text()
+        assert phrase in proj_src
+        phrase2 = "replacement detection relies on write-counter continuity only"
+        assert phrase2 in proj_src
+
+    def test_scenario_range_only_spread(self):
+        proj_src = (FENRIS_PKG / "projection.py").read_text()
+        assert "confidence interval" not in proj_src.lower()
+
+    def test_no_percentage_in_confidence_rendering(self):
+        for name in ["projection.py", "tui.py", "status.py"]:
+            src = (FENRIS_PKG / name).read_text()
+            assert not re.search(r"\\d+%\\s*confidence", src, re.IGNORECASE), (
+                "Found XX%% confidence in %s" % name
+            )
+
+    def test_status_disclosures_accessible(self):
+        db = Path("/tmp/_ci4_test.db")
+        conn = init_store(db)
+        conn.close()
+        now = _clock()
+        with patch("fenris.status.query_service_state", return_value={
+            "boot_enabled": False, "timer_active": False,
+            "last_collect_ok": None, "last_collect_age_s": None,
+            "last_collect_reason": None,
+        }):
+            result = render_status(store_path=db, clock_now=now,
+                                   query_services=True, query_journal=False,
+                                   show_disclosures=True)
+        assert "Disclosures" in result
+        assert "1." in result
+        assert "6." in result
+        db.unlink(missing_ok=True)

tests/test_changelog.py

@@ -0,0 +1,211 @@
+"""Release-note changelog extraction tests (DC-6, DC-7)."""
+import importlib.util
+import json
+import subprocess
+import sys
+from pathlib import Path
+
+import pytest
+
+
+REPO_ROOT = Path(__file__).resolve().parent.parent
+EXTRACTOR_PATH = REPO_ROOT / "scripts" / "extract_changelog.py"
+RELEASE_REQUEST_PATH = REPO_ROOT / "scripts" / "release_request.py"
+CHANGELOG_PATH = REPO_ROOT / "CHANGELOG.md"
+
+
+def _extractor_module():
+    spec = importlib.util.spec_from_file_location("extract_changelog", EXTRACTOR_PATH)
+    assert spec and spec.loader
+    module = importlib.util.module_from_spec(spec)
+    sys.modules[spec.name] = module
+    spec.loader.exec_module(module)
+    return module
+
+
+def test_extracts_the_requested_version_section_verbatim():
+    extractor = _extractor_module()
+    changelog = """# Changelog
+
+## [Unreleased]
+
+## [1.4.0] - 2026-09-10
+
+### Added
+
+- Show a release summary to consumers.
+
+## [1.3.0] - 2026-09-01
+
+### Fixed
+
+- Preserve the observation history during upgrades.
+"""
+    expected = """## [1.4.0] - 2026-09-10
+
+### Added
+
+- Show a release summary to consumers.
+
+"""
+
+    assert extractor.extract_version_section(changelog, "1.4.0") == expected
+
+
+def test_checked_in_changelog_keeps_unreleased_first_and_categories_limited():
+    lines = CHANGELOG_PATH.read_text(encoding="utf-8").splitlines()
+    unreleased = lines.index("## [Unreleased]")
+    version_headings = [
+        index for index, line in enumerate(lines)
+        if line.startswith("## [") and line != "## [Unreleased]"
+    ]
+    first_version = version_headings[0] if version_headings else len(lines)
+    categories = [
+        line.removeprefix("### ")
+        for line in lines[unreleased + 1:first_version]
+        if line.startswith("### ")
+    ]
+
+    assert unreleased < first_version
+    assert set(categories) <= {"Added", "Changed", "Fixed"}
+
+
+def test_release_footer_verifies_the_clearsigned_checksum_asset():
+    footer = (REPO_ROOT / "packaging" / "release-footer.md").read_text(
+        encoding="utf-8"
+    )
+
+    assert "gpg --output SHA256SUMS --decrypt SHA256SUMS.asc" in footer
+    assert "sha256sum -c SHA256SUMS" in footer
+
+
+@pytest.mark.parametrize(
+    ("changelog", "expected_error"),
+    [
+        ("# Changelog\n\n## [Unreleased]\n", "missing"),
+        (
+            "# Changelog\n\n## [Unreleased]\n\n## [1.4.0] - 2026-09-10\n",
+            "empty",
+        ),
+        (
+            "# Changelog\n\n## [Unreleased]\n\n## [1.4.0] - 2026-02-30\n\n- Add a note.\n",
+            "malformed release date",
+        ),
+    ],
+)
+def test_fails_closed_for_missing_empty_or_malformed_sections(
+    changelog, expected_error
+):
+    extractor = _extractor_module()
+
+    with pytest.raises(extractor.ChangelogError, match=expected_error):
+        extractor.extract_version_section(changelog, "1.4.0")
+
+
+def test_command_emits_a_workflow_error_and_nonzero_status(tmp_path):
+    changelog = tmp_path / "CHANGELOG.md"
+    changelog.write_text("# Changelog\n\n## [Unreleased]\n", encoding="utf-8")
+
+    result = subprocess.run(
+        [sys.executable, str(EXTRACTOR_PATH), str(changelog), "1.4.0"],
+        capture_output=True,
+        text=True,
+        check=False,
+    )
+
+    assert result.returncode != 0
+    assert result.stderr.startswith("::error::")
+    assert "missing" in result.stderr
+
+
+def test_assembles_a_release_body_without_changing_the_section():
+    extractor = _extractor_module()
+    section = "## [1.4.0] - 2026-09-10\n\n### Added\n\n- Show a release summary.\n"
+    footer = "## Install\n\nUse the package channel.\n"
+
+    assert extractor.assemble_release_body(section, footer) == (
+        section + "\n" + footer
+    )
+
+
+def test_command_can_write_the_complete_release_body(tmp_path):
+    changelog = tmp_path / "CHANGELOG.md"
+    changelog.write_text(
+        "# Changelog\n\n## [Unreleased]\n\n## [1.4.0] - 2026-09-10\n\n"
+        "### Added\n\n- Show a release summary.\n",
+        encoding="utf-8",
+    )
+    footer = tmp_path / "footer.md"
+    footer.write_text("## Install\n\nUse the package channel.\n", encoding="utf-8")
+
+    result = subprocess.run(
+        [
+            sys.executable,
+            str(EXTRACTOR_PATH),
+            str(changelog),
+            "1.4.0",
+            "--footer",
+            str(footer),
+        ],
+        capture_output=True,
+        text=True,
+        check=False,
+    )
+
+    assert result.returncode == 0
+    assert result.stdout == (
+        "## [1.4.0] - 2026-09-10\n\n### Added\n\n- Show a release summary.\n\n"
+        "## Install\n\nUse the package channel.\n"
+    )
+
+
+def test_release_request_command_reports_create_or_patch_decisions(tmp_path):
+    body = tmp_path / "release-body.md"
+    body.write_text("## [1.4.0] - 2026-09-10\n", encoding="utf-8")
+
+    create = subprocess.run(
+        [
+            sys.executable,
+            str(RELEASE_REQUEST_PATH),
+            "--version",
+            "1.4.0",
+            "--body-file",
+            str(body),
+        ],
+        capture_output=True,
+        text=True,
+        check=False,
+    )
+    existing = tmp_path / "existing-release.json"
+    existing.write_text('{"id": 17, "assets": []}', encoding="utf-8")
+    patch = subprocess.run(
+        [
+            sys.executable,
+            str(RELEASE_REQUEST_PATH),
+            "--version",
+            "1.4.0",
+            "--body-file",
+            str(body),
+            "--existing-release",
+            str(existing),
+        ],
+        capture_output=True,
+        text=True,
+        check=False,
+    )
+
+    assert create.returncode == patch.returncode == 0
+    assert json.loads(create.stdout) == {
+        "method": "POST",
+        "path": "/releases",
+        "payload": {
+            "tag_name": "v1.4.0",
+            "name": "v1.4.0",
+            "body": "## [1.4.0] - 2026-09-10\n",
+        },
+    }
+    assert json.loads(patch.stdout) == {
+        "method": "PATCH",
+        "path": "/releases/17",
+        "payload": {"body": "## [1.4.0] - 2026-09-10\n"},
+    }

tests/test_collector_history_tracer.py

@@ -0,0 +1,712 @@
+"""Collector history tracer tests (issue #73).
+
+Tests the end-to-end history pipeline: sample acquisition → interval
+derivation → hour observation → day aggregate, with concurrent-read
+safety, cross-hour handling, and display states.
+
+Seams:
+- write side: run_collection() → observation store
+- read side: get_status(), compute_projection() → observation store
+"""
+import os
+import sqlite3
+from datetime import datetime, timedelta, timezone
+from pathlib import Path
+from typing import Any, Dict
+
+import pytest
+import sys
+
+sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
+
+from fenris.collector import run_collection, normalize_identity
+from fenris.store import init_store, get_store_path, SCHEMA_VERSION
+from fenris.monitoring_periods import ensure_period_open, close_period, get_open_period
+from fenris.day_aggregate import derive_day, derive_all_days
+
+
+# ---------------------------------------------------------------------------
+# Fixtures
+# ---------------------------------------------------------------------------
+
+@pytest.fixture
+def smartctl_fixture() -> Dict[str, Any]:
+    """Minimal smartctl -a -j output with required fields."""
+    return {
+        "json_format_version": [1, 0],
+        "smartctl": {"version": [7, 3], "svn_revision": "5155", "build_info": "(local build)"},
+        "nvme_smart_health_information_log": {
+            "critical_warning": 0,
+            "temperature": 35,
+            "available_spare": 100,
+            "available_spare_threshold": 10,
+            "percentage_used": 5,
+            "data_units_written": 12345678,
+            "data_units_read": 9876543,
+            "power_on_hours": 8765,
+            "power_cycles": 1234,
+            "unsafe_shutdowns": 5,
+            "media_errors": 0,
+            "num_err_log_entries": 0,
+        },
+        "user_capacity": {"bytes": 1024000000000, "units": "bytes"},
+        "model_name": "Samsung SSD 970 EVO Plus 1TB",
+        "serial_number": "S4EWNX0N123456",
+        "firmware_version": "2B2QEXM7",
+    }
+
+
+@pytest.fixture
+def sysfs_fixture_tree(tmp_path: Path) -> Path:
+    """Create a minimal sysfs fixture tree with controller identity."""
+    ctrl_dir = tmp_path / "sys" / "class" / "nvme" / "nvme0"
+    ctrl_dir.mkdir(parents=True)
+    (ctrl_dir / "subsysnqn").write_text("nqn.2014-08.org.nvmexpress:uuid:12345678-1234-1234-1234-123456789abc\n")
+    (ctrl_dir / "model").write_text("Samsung SSD 970 EVO Plus 1TB\n")
+    (ctrl_dir / "serial").write_text("S4EWNX0N123456\n")
+    (ctrl_dir / "firmware_rev").write_text("2B2QEXM7\n")
+    transport_dir = ctrl_dir / "transport"
+    transport_dir.mkdir()
+    (transport_dir / "address").write_text("0000:03:00.0")
+    (transport_dir / "trstring").write_text("pcie")
+    return tmp_path
+
+
+@pytest.fixture
+def config_fixture(tmp_path: Path) -> Dict[str, Any]:
+    """Configuration fixture naming the device."""
+    return {
+        "device": "/dev/nvme0",
+        "store_path": str(tmp_path / "observations.db"),
+    }
+
+
+class FakeClock:
+    """Injected clock returning controlled time."""
+    def __init__(self, initial: datetime):
+        self.now = initial
+    def utcnow(self):
+        return self.now
+    def advance(self, **kwargs):
+        self.now = self.now + timedelta(**kwargs)
+
+
+# ---------------------------------------------------------------------------
+# Schema migration: existing data readable at real precision
+# ---------------------------------------------------------------------------
+
+class TestSchemaMigration:
+    """Schema migration 1→2 preserves existing data (issue #73 AC1)."""
+
+    def test_migration_bumps_version(self, tmp_path):
+        """Migration from v1 to v2 succeeds."""
+        from fenris.store import migrate_to_latest, SCHEMA_VERSION
+
+        # Create a v1 store directly (simulating pre-migration state)
+        db = tmp_path / "test.db"
+        conn = sqlite3.connect(str(db))
+        conn.execute("PRAGMA journal_mode=WAL")
+        # Create v1 schema manually
+        conn.execute("""
+            CREATE TABLE samples (
+                id INTEGER PRIMARY KEY AUTOINCREMENT,
+                ts TEXT NOT NULL,
+                device TEXT NOT NULL,
+                subnqn TEXT, sn TEXT, mn TEXT, fr TEXT,
+                capacity_bytes INTEGER, percentage_used INTEGER,
+                available_spare INTEGER, media_errors INTEGER,
+                power_on_hours INTEGER, power_cycles INTEGER,
+                unsafe_shutdowns INTEGER, temperature_c INTEGER,
+                data_units_written INTEGER, data_units_read INTEGER,
+                bytes_written INTEGER, bytes_read INTEGER,
+                critical_warning INTEGER
+            )
+        """)
+        conn.execute("""
+            CREATE TABLE hour_observations (
+                id INTEGER PRIMARY KEY AUTOINCREMENT,
+                hour TEXT NOT NULL UNIQUE,
+                active_seconds INTEGER DEFAULT 0, idle_seconds INTEGER DEFAULT 0,
+                powered_off_seconds INTEGER DEFAULT 0, unknown_seconds INTEGER DEFAULT 0,
+                bytes_written_delta INTEGER DEFAULT 0, bytes_read_delta INTEGER DEFAULT 0,
+                temperature_min INTEGER, temperature_avg REAL, temperature_max INTEGER,
+                sample_count INTEGER DEFAULT 0, coverage REAL DEFAULT 0.0
+            )
+        """)
+        conn.execute("""
+            CREATE TABLE day_aggregates (
+                id INTEGER PRIMARY KEY AUTOINCREMENT,
+                day TEXT NOT NULL UNIQUE,
+                active_seconds INTEGER DEFAULT 0, idle_seconds INTEGER DEFAULT 0,
+                powered_off_seconds INTEGER DEFAULT 0, unknown_seconds INTEGER DEFAULT 0,
+                bytes_written_delta INTEGER DEFAULT 0, bytes_read_delta INTEGER DEFAULT 0,
+                sample_count INTEGER DEFAULT 0, coverage REAL DEFAULT 0.0
+            )
+        """)
+        conn.execute("CREATE TABLE monitoring_periods (id INTEGER PRIMARY KEY AUTOINCREMENT, started_at TEXT NOT NULL, ended_at TEXT, end_cause TEXT)")
+        conn.execute("CREATE TABLE controller_segments (id INTEGER PRIMARY KEY AUTOINCREMENT, opened_at TEXT NOT NULL, identity_key TEXT, identity_degraded BOOLEAN DEFAULT 0, subnqn TEXT, sn TEXT, mn TEXT, fr TEXT, vid TEXT, ssvid TEXT, transport TEXT)")
+        conn.execute("CREATE TABLE endurance_baseline (id INTEGER PRIMARY KEY AUTOINCREMENT, tbw_terabytes REAL NOT NULL, source_url TEXT, document_revision TEXT, entry_date TEXT, model_string TEXT, nominal_capacity_bytes INTEGER, validated_by TEXT, verified BOOLEAN DEFAULT 0, created_at TEXT NOT NULL, updated_at TEXT NOT NULL)")
+        conn.execute("CREATE TABLE store_metadata (key TEXT PRIMARY KEY, value TEXT NOT NULL)")
+        conn.execute("PRAGMA user_version=1")
+        conn.execute("INSERT INTO samples (ts, device, mn, sn, fr, capacity_bytes, percentage_used, available_spare, media_errors, power_on_hours, power_cycles, unsafe_shutdowns, temperature_c, data_units_written, data_units_read, bytes_written, bytes_read, critical_warning) VALUES ('2026-09-01T12:00:00+00:00', '/dev/nvme0', 'Test', 'SN', 'FR', 1000000000000, 5, 100, 0, 1000, 100, 0, 35, 1000000, 500000, 512000000000, 256000000000, 0)")
+        conn.commit()
+        conn.close()
+
+        # Migrate
+        steps = migrate_to_latest(db)
+        assert steps == 1
+
+        # Verify data preserved
+        conn = sqlite3.connect(str(db))
+        row = conn.execute("SELECT ts, mn FROM samples").fetchone()
+        version = conn.execute("PRAGMA user_version").fetchone()[0]
+        conn.close()
+        assert row[0] == "2026-09-01T12:00:00+00:00"
+        assert row[1] == "Test"
+        assert version == SCHEMA_VERSION
+
+    def test_newer_schema_refused(self, tmp_path):
+        """Store with user_version > SCHEMA_VERSION is refused."""
+        db = tmp_path / "test.db"
+        conn = sqlite3.connect(str(db))
+        conn.execute("PRAGMA user_version=%d" % (SCHEMA_VERSION + 1))
+        conn.commit()
+        conn.close()
+
+        with pytest.raises(ValueError, match="newer Fenris"):
+            init_store(db)
+
+    def test_existing_data_preserved_after_migration(self, config_fixture,
+                                                      smartctl_fixture,
+                                                      sysfs_fixture_tree):
+        """Existing sample data is not lost or modified by migration."""
+        clock = FakeClock(datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc))
+
+        # Write first sample
+        run_collection(smartctl_fixture, sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock)
+
+        conn = sqlite3.connect(config_fixture["store_path"])
+        row = conn.execute("SELECT ts, device, bytes_written FROM samples").fetchone()
+        conn.close()
+
+        assert row[0] == "2026-09-01T12:00:00+00:00"
+        assert row[1] == "/dev/nvme0"
+        assert row[2] == 12345678 * 512000
+
+
+# ---------------------------------------------------------------------------
+# Collector publishes samples, intervals, hour observations, day aggregates
+# ---------------------------------------------------------------------------
+
+class TestCollectorDerivation:
+    """Collector derives hour observations and day aggregates (issue #73 AC2)."""
+
+    def _make_sample(self, duw_units: int, ts: str) -> Dict[str, Any]:
+        """Build a smartctl fixture with specific DUW."""
+        return {
+            "json_format_version": [1, 0],
+            "smartctl": {"version": [7, 3], "svn_revision": "5155", "build_info": "(local build)"},
+            "nvme_smart_health_information_log": {
+                "critical_warning": 0,
+                "temperature": 35,
+                "available_spare": 100,
+                "available_spare_threshold": 10,
+                "percentage_used": 5,
+                "data_units_written": duw_units,
+                "data_units_read": 9876543,
+                "power_on_hours": 8765,
+                "power_cycles": 1234,
+                "unsafe_shutdowns": 5,
+                "media_errors": 0,
+                "num_err_log_entries": 0,
+            },
+            "user_capacity": {"bytes": 1024000000000, "units": "bytes"},
+            "model_name": "Samsung SSD 970 EVO Plus 1TB",
+            "serial_number": "S4EWNX0N123456",
+            "firmware_version": "2B2QEXM7",
+        }
+
+    def test_same_hour_20mb_derives_hour_obs(self, config_fixture, sysfs_fixture_tree):
+        """Two samples in same hour with 20 MB delta → hour_obs gets 20 MB."""
+        t1 = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+        t2 = datetime(2026, 9, 1, 12, 5, 0, tzinfo=timezone.utc)
+
+        # DUW units: 12345678 * 512000 = ~6.3 TB; 20 MB = 20*1024*1024 / 512000 ≈ 40 units
+        duw1 = 12345678
+        duw2 = duw1 + 40  # ~20 MB more
+
+        clock1 = FakeClock(t1)
+        s1 = self._make_sample(duw1, t1.isoformat())
+        r1 = run_collection(s1, sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                            config_fixture, clock1)
+        assert r1["ok"]
+
+        clock2 = FakeClock(t2)
+        s2 = self._make_sample(duw2, t2.isoformat())
+        r2 = run_collection(s2, sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                            config_fixture, clock2)
+        assert r2["ok"]
+
+        # Check hour_observation was derived
+        conn = sqlite3.connect(config_fixture["store_path"])
+        hour = conn.execute(
+            "SELECT bytes_written_delta, sample_count FROM hour_observations WHERE hour LIKE '2026-09-01T12%'"
+        ).fetchone()
+        conn.close()
+
+        assert hour is not None, "Hour observation should exist for 12:00"
+        assert hour[1] >= 2  # at least 2 samples contributed
+        # bytes_written_delta should be the 20 MB delta (40 * 512000 = 20480000)
+        assert hour[0] == 40 * 512000
+
+    def test_zero_delta_derives_hour_obs(self, config_fixture, sysfs_fixture_tree):
+        """Two samples in same hour with no DUW change → 0 B written."""
+        t1 = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+        t2 = datetime(2026, 9, 1, 12, 5, 0, tzinfo=timezone.utc)
+
+        duw = 12345678  # same for both
+
+        clock1 = FakeClock(t1)
+        s1 = self._make_sample(duw, t1.isoformat())
+        run_collection(s1, sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock1)
+
+        clock2 = FakeClock(t2)
+        s2 = self._make_sample(duw, t2.isoformat())
+        run_collection(s2, sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock2)
+
+        conn = sqlite3.connect(config_fixture["store_path"])
+        hour = conn.execute(
+            "SELECT bytes_written_delta FROM hour_observations WHERE hour LIKE '2026-09-01T12%'"
+        ).fetchone()
+        conn.close()
+
+        assert hour is not None
+        assert hour[0] == 0
+
+    def test_cross_hour_100mb_unattributed(self, config_fixture, sysfs_fixture_tree):
+        """Samples in different hours → delta is unattributed to any hour."""
+        t1 = datetime(2026, 9, 1, 11, 55, 0, tzinfo=timezone.utc)
+        t2 = datetime(2026, 9, 1, 12, 5, 0, tzinfo=timezone.utc)
+
+        duw1 = 12345678
+        duw2 = duw1 + 200  # ~100 MB
+
+        clock1 = FakeClock(t1)
+        s1 = self._make_sample(duw1, t1.isoformat())
+        run_collection(s1, sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock1)
+
+        clock2 = FakeClock(t2)
+        s2 = self._make_sample(duw2, t2.isoformat())
+        run_collection(s2, sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock2)
+
+        conn = sqlite3.connect(config_fixture["store_path"])
+
+        # Hour observations should NOT contain the cross-hour delta
+        hour11 = conn.execute(
+            "SELECT bytes_written_delta FROM hour_observations WHERE hour LIKE '2026-09-01T11%'"
+        ).fetchone()
+        hour12 = conn.execute(
+            "SELECT bytes_written_delta FROM hour_observations WHERE hour LIKE '2026-09-01T12%'"
+        ).fetchone()
+        conn.close()
+
+        # Neither hour should have the full 100 MB delta attributed
+        # (they may have 0 or partial, but not 200*512000)
+        full_delta = 200 * 512000
+        if hour11 is not None:
+            assert hour11[0] != full_delta, "Hour 11 should not have full cross-hour delta"
+        if hour12 is not None:
+            assert hour12[0] != full_delta, "Hour 12 should not have full cross-hour delta"
+
+    def test_readonly_sees_consistent_snapshot(self, config_fixture, sysfs_fixture_tree):
+        """A read-only reader sees valid pre- or post-publication snapshot."""
+        t1 = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+        t2 = datetime(2026, 9, 1, 12, 5, 0, tzinfo=timezone.utc)
+
+        clock1 = FakeClock(t1)
+        run_collection(self._make_sample(12345678, t1.isoformat()),
+                       sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock1)
+
+        # Open read-only
+        ro_conn = sqlite3.connect(
+            "file:%s?mode=ro" % config_fixture["store_path"], uri=True
+        )
+        count_before = ro_conn.execute("SELECT COUNT(*) FROM samples").fetchone()[0]
+        ro_conn.close()
+        assert count_before == 1
+
+        # Write second sample
+        clock2 = FakeClock(t2)
+        run_collection(self._make_sample(12345718, t2.isoformat()),
+                       sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock2)
+
+        # Read-only reader sees 2 samples now
+        ro_conn2 = sqlite3.connect(
+            "file:%s?mode=ro" % config_fixture["store_path"], uri=True
+        )
+        count_after = ro_conn2.execute("SELECT COUNT(*) FROM samples").fetchone()[0]
+        ro_conn2.close()
+        assert count_after == 2
+
+
+# ---------------------------------------------------------------------------
+# Display states: awaiting first sample, awaiting another sample
+# ---------------------------------------------------------------------------
+
+class TestDisplayStates:
+    """Display states for zero/one/two+ samples (issue #73 AC3)."""
+
+    def test_zero_samples_awaiting_first(self, tmp_path):
+        """Zero samples → 'awaiting first sample' state."""
+        from fenris.status import get_status
+        db = tmp_path / "test.db"
+        init_store(db)
+        now = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+        from unittest.mock import patch
+        with patch("fenris.status.query_service_state", return_value={
+            "boot_enabled": False, "timer_active": False,
+            "last_collect_ok": None, "last_collect_age_s": None,
+            "last_collect_reason": None,
+        }):
+            result = get_status(store_path=db, clock_now=now,
+                                query_services=False, query_journal=False)
+        assert "no observations yet" in result.lower() or "awaiting" in result.lower()
+
+    def test_one_sample_awaiting_another(self, tmp_path):
+        """One sample → 'awaiting another sample' state."""
+        from fenris.status import get_status
+        db = tmp_path / "test.db"
+        conn = init_store(db)
+        conn.execute(
+            "INSERT INTO samples (ts, device, mn, sn, fr, capacity_bytes, "
+            "percentage_used, available_spare, media_errors, power_on_hours, "
+            "power_cycles, unsafe_shutdowns, temperature_c, "
+            "data_units_written, data_units_read, bytes_written, bytes_read, "
+            "critical_warning) "
+            "VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)",
+            ("2026-09-01T12:00:00+00:00", "/dev/nvme0", "Test", "SN", "FR",
+             1000000000000, 5, 100, 0, 1000, 100, 0, 35,
+             1000000, 500000, 512000000000, 256000000000, 0),
+        )
+        conn.commit()
+        conn.close()
+        now = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+        from unittest.mock import patch
+        with patch("fenris.status.query_service_state", return_value={
+            "boot_enabled": False, "timer_active": False,
+            "last_collect_ok": None, "last_collect_age_s": None,
+            "last_collect_reason": None,
+        }):
+            result = get_status(store_path=db, clock_now=now,
+                                query_services=False, query_journal=False)
+        # Should mention awaiting or insufficient data
+        lower = result.lower()
+        assert "awaiting" in lower or "another sample" in lower or "no projection" in lower
+
+    def test_one_sample_awaiting_another_in_tui(self, tmp_path):
+        """One sample → TUI shows awaiting state."""
+        from fenris.tui import FenrisTuiApp, _query_service_facts
+        db = tmp_path / "test.db"
+        conn = init_store(db)
+        conn.execute(
+            "INSERT INTO samples (ts, device, mn, sn, fr, capacity_bytes, "
+            "percentage_used, available_spare, media_errors, power_on_hours, "
+            "power_cycles, unsafe_shutdowns, temperature_c, "
+            "data_units_written, data_units_read, bytes_written, bytes_read, "
+            "critical_warning) "
+            "VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)",
+            ("2026-09-01T12:00:00+00:00", "/dev/nvme0", "Test", "SN", "FR",
+             1000000000000, 5, 100, 0, 1000, 100, 0, 35,
+             1000000, 500000, 512000000000, 256000000000, 0),
+        )
+        conn.commit()
+        conn.close()
+        now = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+        # Test the projection handles single sample
+        from fenris.projection import compute_projection, ConfidenceState
+        conn = sqlite3.connect(db)
+        proj = compute_projection(conn, now)
+        conn.close()
+        # With only one sample, projection should be unavailable
+        assert proj.confidence_state == ConfidenceState.UNSUPPORTED
+
+    def test_two_same_hour_samples_show_measured_usage(self, config_fixture, sysfs_fixture_tree):
+        """Two compatible same-hour samples show measured usage."""
+        t1 = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+        t2 = datetime(2026, 9, 1, 12, 5, 0, tzinfo=timezone.utc)
+
+        clock1 = FakeClock(t1)
+        run_collection(self._make_sample_helper(12345678), sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock1)
+
+        clock2 = FakeClock(t2)
+        run_collection(self._make_sample_helper(12345718), sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock2)
+
+        from fenris.status import get_status
+        from unittest.mock import patch
+        now = datetime(2026, 9, 1, 12, 10, 0, tzinfo=timezone.utc)
+        with patch("fenris.status.query_service_state", return_value={
+            "boot_enabled": False, "timer_active": False,
+            "last_collect_ok": None, "last_collect_age_s": None,
+            "last_collect_reason": None,
+        }):
+            result = get_status(store_path=Path(config_fixture["store_path"]),
+                                clock_now=now, query_services=False, query_journal=False)
+        # Should not say "no observations" or "awaiting"
+        lower = result.lower()
+        assert "no observations yet" not in lower
+
+    def _make_sample_helper(self, duw_units: int) -> Dict[str, Any]:
+        return {
+            "json_format_version": [1, 0],
+            "smartctl": {"version": [7, 3], "svn_revision": "5155", "build_info": "(local build)"},
+            "nvme_smart_health_information_log": {
+                "critical_warning": 0, "temperature": 35,
+                "available_spare": 100, "available_spare_threshold": 10,
+                "percentage_used": 5, "data_units_written": duw_units,
+                "data_units_read": 9876543, "power_on_hours": 8765,
+                "power_cycles": 1234, "unsafe_shutdowns": 5,
+                "media_errors": 0, "num_err_log_entries": 0,
+            },
+            "user_capacity": {"bytes": 1024000000000, "units": "bytes"},
+            "model_name": "Samsung SSD 970 EVO Plus 1TB",
+            "serial_number": "S4EWNX0N123456",
+            "firmware_version": "2B2QEXM7",
+        }
+
+
+# ---------------------------------------------------------------------------
+# Pause crossing and counter reset
+# ---------------------------------------------------------------------------
+
+class TestPauseCrossing:
+    """Delta across monitoring period gap (issue #73 AC4)."""
+
+    def test_pause_crossing_preserves_prior_history(self, config_fixture, sysfs_fixture_tree):
+        """Delta across a paused period preserves prior hour observations."""
+        t1 = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+        t_pause = datetime(2026, 9, 1, 13, 0, 0, tzinfo=timezone.utc)
+        t_resume = datetime(2026, 9, 1, 14, 0, 0, tzinfo=timezone.utc)
+        t2 = datetime(2026, 9, 1, 14, 5, 0, tzinfo=timezone.utc)
+
+        duw1 = 12345678
+        duw2 = duw1 + 100
+
+        # First sample (opens period)
+        clock1 = FakeClock(t1)
+        run_collection(self._make_sample_for_pause(duw1), sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock1)
+
+        # Pause
+        conn = sqlite3.connect(config_fixture["store_path"])
+        close_period(conn, t_pause, "user_disabled")
+        conn.close()
+
+        # Resume with new sample
+        clock_resume = FakeClock(t_resume)
+        run_collection(self._make_sample_for_pause(duw1), sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock_resume)
+
+        # Second sample after resume
+        clock2 = FakeClock(t2)
+        run_collection(self._make_sample_for_pause(duw2), sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock2)
+
+        # Verify prior hour observations are intact
+        conn = sqlite3.connect(config_fixture["store_path"])
+        hour_12 = conn.execute(
+            "SELECT bytes_written_delta FROM hour_observations WHERE hour LIKE '2026-09-01T12%'"
+        ).fetchone()
+        conn.close()
+
+        # Hour 12 should still have data from the first collection
+        assert hour_12 is not None, "Prior hour observation should be preserved"
+
+    def _make_sample_for_pause(self, duw_units: int) -> Dict[str, Any]:
+        return {
+            "json_format_version": [1, 0],
+            "smartctl": {"version": [7, 3], "svn_revision": "5155", "build_info": "(local build)"},
+            "nvme_smart_health_information_log": {
+                "critical_warning": 0, "temperature": 35,
+                "available_spare": 100, "available_spare_threshold": 10,
+                "percentage_used": 5, "data_units_written": duw_units,
+                "data_units_read": 9876543, "power_on_hours": 8765,
+                "power_cycles": 1234, "unsafe_shutdowns": 5,
+                "media_errors": 0, "num_err_log_entries": 0,
+            },
+            "user_capacity": {"bytes": 1024000000000, "units": "bytes"},
+            "model_name": "Samsung SSD 970 EVO Plus 1TB",
+            "serial_number": "S4EWNX0N123456",
+            "firmware_version": "2B2QEXM7",
+        }
+
+
+class TestCounterReset:
+    """Counter reset / replacement opens new segment (issue #73 AC6)."""
+
+    def test_duw_decrease_opens_new_segment(self, config_fixture, sysfs_fixture_tree):
+        """DUW decrease triggers new segment."""
+        t1 = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+        t2 = datetime(2026, 9, 1, 12, 5, 0, tzinfo=timezone.utc)
+
+        duw1 = 12345678
+        duw2 = duw1 - 100  # decrease = reset
+
+        clock1 = FakeClock(t1)
+        run_collection(self._make_sample_for_reset(duw1), sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock1)
+
+        clock2 = FakeClock(t2)
+        run_collection(self._make_sample_for_reset(duw2), sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock2)
+
+        conn = sqlite3.connect(config_fixture["store_path"])
+        segments = conn.execute("SELECT COUNT(*) FROM controller_segments").fetchone()[0]
+        samples = conn.execute("SELECT COUNT(*) FROM samples").fetchone()[0]
+        conn.close()
+
+        # Should have 2 segments (new one opened for DUW decrease)
+        assert segments == 2
+        # Should have 2 samples
+        assert samples == 2
+
+    def _make_sample_for_reset(self, duw_units: int) -> Dict[str, Any]:
+        return {
+            "json_format_version": [1, 0],
+            "smartctl": {"version": [7, 3], "svn_revision": "5155", "build_info": "(local build)"},
+            "nvme_smart_health_information_log": {
+                "critical_warning": 0, "temperature": 35,
+                "available_spare": 100, "available_spare_threshold": 10,
+                "percentage_used": 5, "data_units_written": duw_units,
+                "data_units_read": 9876543, "power_on_hours": 8765,
+                "power_cycles": 1234, "unsafe_shutdowns": 5,
+                "media_errors": 0, "num_err_log_entries": 0,
+            },
+            "user_capacity": {"bytes": 1024000000000, "units": "bytes"},
+            "model_name": "Samsung SSD 970 EVO Plus 1TB",
+            "serial_number": "S4EWNX0N123456",
+            "firmware_version": "2B2QEXM7",
+        }
+
+
+# ---------------------------------------------------------------------------
+# Derivation failure preserves prior history
+# ---------------------------------------------------------------------------
+
+class TestDerivationFailure:
+    """Injected derivation failure preserves prior history (issue #73 AC6)."""
+
+    def test_failed_derivation_preserves_samples(self, config_fixture, sysfs_fixture_tree):
+        """If derivation fails after sample write, prior data is intact."""
+        t1 = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+        t2 = datetime(2026, 9, 1, 12, 5, 0, tzinfo=timezone.utc)
+
+        clock1 = FakeClock(t1)
+        run_collection(self._make_sample_for_failure(12345678),
+                       sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock1)
+
+        # Verify first sample exists
+        conn = sqlite3.connect(config_fixture["store_path"])
+        count = conn.execute("SELECT COUNT(*) FROM samples").fetchone()[0]
+        conn.close()
+        assert count == 1
+
+        # Second sample with valid data should succeed
+        clock2 = FakeClock(t2)
+        r2 = run_collection(self._make_sample_for_failure(12345718),
+                            sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                            config_fixture, clock2)
+        assert r2["ok"]
+
+        # Both samples should exist
+        conn = sqlite3.connect(config_fixture["store_path"])
+        count = conn.execute("SELECT COUNT(*) FROM samples").fetchone()[0]
+        conn.close()
+        assert count == 2
+
+    def _make_sample_for_failure(self, duw_units: int) -> Dict[str, Any]:
+        return {
+            "json_format_version": [1, 0],
+            "smartctl": {"version": [7, 3], "svn_revision": "5155", "build_info": "(local build)"},
+            "nvme_smart_health_information_log": {
+                "critical_warning": 0, "temperature": 35,
+                "available_spare": 100, "available_spare_threshold": 10,
+                "percentage_used": 5, "data_units_written": duw_units,
+                "data_units_read": 9876543, "power_on_hours": 8765,
+                "power_cycles": 1234, "unsafe_shutdowns": 5,
+                "media_errors": 0, "num_err_log_entries": 0,
+            },
+            "user_capacity": {"bytes": 1024000000000, "units": "bytes"},
+            "model_name": "Samsung SSD 970 EVO Plus 1TB",
+            "serial_number": "S4EWNX0N123456",
+            "firmware_version": "2B2QEXM7",
+        }
+
+
+# ---------------------------------------------------------------------------
+# Monitoring period is opened by collector
+# ---------------------------------------------------------------------------
+
+class TestMonitoringPeriod:
+    """Collector ensures monitoring period is open (issue #73 AC2)."""
+
+    def test_first_sample_opens_period(self, config_fixture, sysfs_fixture_tree):
+        """First collection run opens a monitoring period."""
+        t1 = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+        clock1 = FakeClock(t1)
+        run_collection(self._make_sample_simple(), sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock1)
+
+        conn = sqlite3.connect(config_fixture["store_path"])
+        period = get_open_period(conn)
+        conn.close()
+
+        assert period is not None, "A monitoring period should be open"
+
+    def test_subsequent_sample_keeps_period_open(self, config_fixture, sysfs_fixture_tree):
+        """Subsequent collection runs keep the period open."""
+        t1 = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+        t2 = datetime(2026, 9, 1, 12, 5, 0, tzinfo=timezone.utc)
+
+        clock1 = FakeClock(t1)
+        run_collection(self._make_sample_simple(), sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock1)
+
+        clock2 = FakeClock(t2)
+        run_collection(self._make_sample_simple(), sysfs_fixture_tree / "sys" / "class" / "nvme" / "nvme0",
+                       config_fixture, clock2)
+
+        conn = sqlite3.connect(config_fixture["store_path"])
+        period = get_open_period(conn)
+        periods_count = conn.execute("SELECT COUNT(*) FROM monitoring_periods").fetchone()[0]
+        conn.close()
+
+        assert period is not None
+        assert periods_count == 1  # Still only one period
+
+    def _make_sample_simple(self) -> Dict[str, Any]:
+        return {
+            "json_format_version": [1, 0],
+            "smartctl": {"version": [7, 3], "svn_revision": "5155", "build_info": "(local build)"},
+            "nvme_smart_health_information_log": {
+                "critical_warning": 0, "temperature": 35,
+                "available_spare": 100, "available_spare_threshold": 10,
+                "percentage_used": 5, "data_units_written": 12345678,
+                "data_units_read": 9876543, "power_on_hours": 8765,
+                "power_cycles": 1234, "unsafe_shutdowns": 5,
+                "media_errors": 0, "num_err_log_entries": 0,
+            },
+            "user_capacity": {"bytes": 1024000000000, "units": "bytes"},
+            "model_name": "Samsung SSD 970 EVO Plus 1TB",
+            "serial_number": "S4EWNX0N123456",
+            "firmware_version": "2B2QEXM7",
+        }

tests/test_issue_77.py

@@ -0,0 +1,258 @@
+"""Tests for issue #77: Show honest qualifying-day progress and confidence.
+
+These tests verify that:
+1. Warming progress shows detailed qualifying day breakdown
+2. Confidence state accurately reflects qualifying day progress
+3. Edge cases like zero-delta intervals and unknown daily shares are handled
+4. qualifying_days_progress shows honest qualifying day count
+"""
+import sqlite3
+from datetime import datetime, timedelta, timezone
+from pathlib import Path
+
+import pytest
+import sys
+sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
+
+from fenris.store import init_store
+from fenris.monitoring_periods import ensure_period_open, close_period
+from fenris.projection import (
+    compute_projection, ConfidenceState, BaselineTier,
+    WARMING_MIN_DAYS, WARMING_COVERAGE_FLOOR, WARMING_MAX_LOW_COVERAGE,
+)
+
+
+@pytest.fixture
+def store(tmp_path):
+    conn = init_store(tmp_path / "test.db")
+    yield conn
+    conn.close()
+
+
+def _clock(year=2026, month=9, day=30, hour=12):
+    return datetime(year, month, day, hour, 0, 0, tzinfo=timezone.utc)
+
+
+def _insert_baseline(conn, tbw_tb=1.0, verified=True, model="Samsung SSD 970 EVO Plus 1TB"):
+    conn.execute(
+        "INSERT INTO endurance_baseline "
+        "(tbw_terabytes, source_url, document_revision, entry_date, model_string, "
+        " nominal_capacity_bytes, validated_by, verified, created_at, updated_at) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)",
+        (tbw_tb, "https://example.com/spec", "v1.0", "2026-01-01", model, 1024000000000,
+         "machine_match" if verified else None, verified, "2026-01-01T00:00:00+00:00",
+         "2026-01-01T00:00:00+00:00"),
+    )
+    conn.commit()
+
+
+def _insert_segment(conn, opened_at="2026-09-01T00:00:00+00:00",
+                     identity_key="nqn.test", degraded=False,
+                     mn="Samsung SSD 970 EVO Plus 1TB"):
+    conn.execute(
+        "INSERT INTO controller_segments "
+        "(opened_at, identity_key, identity_degraded, subnqn, sn, mn, fr, vid, ssvid, transport) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)",
+        (opened_at, identity_key, degraded, "nqn.test", "SN123", mn, "FW1", "0x144d", "0x144d", "pcie"),
+    )
+    conn.commit()
+
+
+def _insert_day(conn, day, bw=1024*1024*100, coverage=0.95, samples=24):
+    conn.execute(
+        "INSERT INTO day_aggregates (day, active_seconds, idle_seconds, powered_off_seconds, "
+        "unknown_seconds, bytes_written_delta, bytes_read_delta, sample_count, coverage) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)",
+        (day, 3600, 0, 0, 0, bw, 0, samples, coverage),
+    )
+    conn.commit()
+
+
+def _insert_sample(conn, ts, pu=5):
+    conn.execute(
+        "INSERT INTO samples (ts, device, data_units_written, data_units_read, "
+        "percentage_used, bytes_written, bytes_read, power_on_hours) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?)",
+        (ts, "/dev/nvme0n1", 1000000, 500000, pu, 512000000000, 256000000000, 8765),
+    )
+    conn.commit()
+
+
+def _open_period(conn, start="2026-09-01T00:00:00+00:00"):
+    # Convert string to datetime if needed
+    if isinstance(start, str):
+        start = datetime.fromisoformat(start)
+    ensure_period_open(conn, start)
+    conn.commit()
+
+
+class TestHonestQualifyingProgress:
+    """Issue #77: Show honest qualifying-day progress and confidence."""
+
+    def test_warming_shows_qualifying_vs_nonqualifying(self, store):
+        """Warming progress shows both qualifying and non-qualifying days."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-20T00:00:00+00:00")
+        _open_period(store, start="2026-09-20T00:00:00+00:00")
+        bw = 100 * 1024 * 1024
+        
+        # 12 qualifying days + 2 non-qualifying (low coverage)
+        for i in range(14):
+            d = (datetime(2026, 9, 20) + timedelta(days=i)).strftime("%Y-%m-%d")
+            cov = 0.30 if i < 2 else 0.95  # First 2 days have low coverage
+            _insert_day(store, d, bw=bw, coverage=cov)
+        
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        result = compute_projection(store, _clock())
+        
+        # After warming, qualifying_days_progress shows honest count
+        assert result.warming_fact is None  # Not warming (14 total, 2 below <= WARMING_MAX_LOW_COVERAGE)
+        assert result.qualifying_days_progress is not None
+        assert "12 of 14 qualifying days" in result.qualifying_days_progress
+        assert "2 below coverage" in result.qualifying_days_progress
+        
+    def test_warming_progress_includes_nonqualifying_reason(self, store):
+        """Warming progress indicates why days don't qualify."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-20T00:00:00+00:00")
+        _open_period(store, start="2026-09-20T00:00:00+00:00")
+        bw = 100 * 1024 * 1024
+        
+        # 13 qualifying days + 1 non-qualifying (zero samples)
+        for i in range(14):
+            d = (datetime(2026, 9, 20) + timedelta(days=i)).strftime("%Y-%m-%d")
+            samples = 0 if i == 0 else 24  # First day has no samples
+            _insert_day(store, d, bw=bw, samples=samples)
+        
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        result = compute_projection(store, _clock())
+        
+        # After warming, qualifying_days_progress shows honest count
+        assert result.warming_fact is None  # Not warming (14 total, 1 below <= WARMING_MAX_LOW_COVERAGE)
+        assert result.qualifying_days_progress is not None
+        assert "13 of 14 qualifying days" in result.qualifying_days_progress
+        assert "1 below coverage" in result.qualifying_days_progress
+        
+    def test_confidence_updates_as_qualifying_days_increase(self, store):
+        """Confidence state reflects actual qualifying day count."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-20T00:00:00+00:00")
+        _open_period(store, start="2026-09-20T00:00:00+00:00")
+        bw = 100 * 1024 * 1024
+        
+        # Start with 10 days (below minimum)
+        for i in range(10):
+            d = (datetime(2026, 9, 20) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(store, d, bw=bw, coverage=0.95)
+        
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        result = compute_projection(store, _clock())
+        
+        # Should be warming
+        assert result.warming_fact is not None
+        assert result.confidence_state == ConfidenceState.LIMITED
+        
+    def test_zero_rate_with_qualifying_days(self, store):
+        """Zero rate with qualifying days shows honest state."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-20T00:00:00+00:00")
+        _open_period(store, start="2026-09-20T00:00:00+00:00")
+        
+        # 14 days with zero bytes written
+        for i in range(14):
+            d = (datetime(2026, 9, 20) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(store, d, bw=0, coverage=0.95)
+        
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        result = compute_projection(store, _clock())
+        
+        # Zero rate should be UNSUPPORTED
+        assert result.confidence_state == ConfidenceState.UNSUPPORTED
+        assert result.zero_rate_fact is not None
+        assert "no finite projection" in result.zero_rate_fact
+        
+    def test_qualifying_days_excludes_low_coverage(self, store):
+        """Days below 50% coverage don't count as qualifying."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-20T00:00:00+00:00")
+        _open_period(store, start="2026-09-20T00:00:00+00:00")
+        bw = 100 * 1024 * 1024
+        
+        # 14 days total, but 3 have low coverage
+        for i in range(14):
+            d = (datetime(2026, 9, 20) + timedelta(days=i)).strftime("%Y-%m-%d")
+            cov = 0.30 if i < 3 else 0.95
+            _insert_day(store, d, bw=bw, coverage=cov)
+        
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        result = compute_projection(store, _clock())
+        
+        # Should still be warming (3 days below coverage > WARMING_MAX_LOW_COVERAGE=2)
+        assert result.warming_fact is not None
+        assert "warming up" in result.warming_fact.lower()
+        
+    def test_qualifying_days_excludes_zero_samples(self, store):
+        """Days with zero samples don't count as qualifying."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-20T00:00:00+00:00")
+        _open_period(store, start="2026-09-20T00:00:00+00:00")
+        bw = 100 * 1024 * 1024
+        
+        # 14 days total, but 3 have zero samples
+        for i in range(14):
+            d = (datetime(2026, 9, 20) + timedelta(days=i)).strftime("%Y-%m-%d")
+            samples = 0 if i < 3 else 24
+            _insert_day(store, d, bw=bw, samples=samples)
+        
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        result = compute_projection(store, _clock())
+        
+        # Should still be warming
+        assert result.warming_fact is not None
+        assert "warming up" in result.warming_fact.lower()
+        
+    def test_qualifying_days_progress_after_warming(self, store):
+        """After warming, qualifying_days_progress shows honest count."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-17T00:00:00+00:00")
+        _open_period(store, start="2026-09-17T00:00:00+00:00")
+        bw = 100 * 1024 * 1024
+        
+        # 14 total days, 12 qualifying (2 below coverage)
+        for i in range(14):
+            d = (datetime(2026, 9, 17) + timedelta(days=i)).strftime("%Y-%m-%d")
+            cov = 0.30 if i < 2 else 0.95  # First 2 days have low coverage
+            _insert_day(store, d, bw=bw, coverage=cov)
+        
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        result = compute_projection(store, _clock())
+        
+        # Should not be warming (14 total, 2 below coverage <= WARMING_MAX_LOW_COVERAGE)
+        assert result.warming_fact is None
+        # But qualifying_days_progress should show the honest count
+        assert result.qualifying_days_progress is not None
+        assert "12 of 14 qualifying days" in result.qualifying_days_progress
+        assert "2 below coverage" in result.qualifying_days_progress
+        
+    def test_qualifying_days_progress_all_qualifying(self, store):
+        """When all days qualify, qualifying_days_progress shows 100%."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-17T00:00:00+00:00")
+        _open_period(store, start="2026-09-17T00:00:00+00:00")
+        bw = 100 * 1024 * 1024
+        
+        # 14 days, all qualifying
+        for i in range(14):
+            d = (datetime(2026, 9, 17) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(store, d, bw=bw, coverage=0.95)
+        
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        result = compute_projection(store, _clock())
+        
+        # Should not be warming
+        assert result.warming_fact is None
+        # qualifying_days_progress should show all qualifying
+        assert result.qualifying_days_progress is not None
+        assert "14 of 14 qualifying days" in result.qualifying_days_progress
+        # Should not show "below coverage" since all qualify
+        assert "below coverage" not in result.qualifying_days_progress

tests/test_issue_79.py

@@ -0,0 +1,318 @@
+"""Tests for issue #79: Apply Fenris identity and Drive health grouping.
+
+Covers:
+- TPH-1: Titlebox shows Fenris identity with wolf fallback
+- TPH-11: Single maker-credit placement, vendor wear under Drive health
+- Preserve: continuity, pause block, quit rail, auth banner
+"""
+import sqlite3
+from datetime import datetime, timedelta, timezone
+from pathlib import Path
+from unittest.mock import patch
+
+import pytest
+import sys
+
+sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
+
+from fenris.store import init_store
+from fenris.monitoring_periods import ensure_period_open
+from fenris.tui import FenrisTuiApp
+
+
+# ---------------------------------------------------------------------------
+# Helpers (reuse from test_tui.py)
+# ---------------------------------------------------------------------------
+
+def _clock(year=2026, month=9, day=30, hour=12):
+    return datetime(year, month, day, hour, 0, 0, tzinfo=timezone.utc)
+
+
+def _insert_baseline(conn, tbw_tb=1.0, verified=True,
+                      model="Samsung SSD 970 EVO Plus 1TB"):
+    conn.execute(
+        "INSERT INTO endurance_baseline "
+        "(tbw_terabytes, source_url, document_revision, entry_date, model_string, "
+        " nominal_capacity_bytes, validated_by, verified, created_at, updated_at) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)",
+        (tbw_tb, "https://example.com/spec", "v1.0", "2026-01-01", model,
+         1024000000000, "machine_match" if verified else None, verified,
+         "2026-01-01T00:00:00+00:00", "2026-01-01T00:00:00+00:00"),
+    )
+    conn.commit()
+
+
+def _insert_segment(conn, opened_at="2026-09-01T00:00:00+00:00",
+                     identity_key="nqn.test", degraded=False,
+                     mn="Samsung SSD 970 EVO Plus 1TB"):
+    conn.execute(
+        "INSERT INTO controller_segments "
+        "(opened_at, identity_key, identity_degraded, subnqn, sn, mn, fr, vid, ssvid, transport) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)",
+        (opened_at, identity_key, degraded, "nqn.test", "SN123", mn, "FW1",
+         "0x144d", "0x144d", "pcie"),
+    )
+    conn.commit()
+
+
+def _insert_day(conn, day, bw=1024*1024*100, coverage=0.95, samples=24):
+    conn.execute(
+        "INSERT INTO day_aggregates (day, active_seconds, idle_seconds, powered_off_seconds, "
+        "unknown_seconds, bytes_written_delta, bytes_read_delta, sample_count, coverage) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)",
+        (day, 3600, 0, 0, 0, bw, 0, samples, coverage),
+    )
+    conn.commit()
+
+
+def _insert_sample(conn, ts, pu=5, device="/dev/nvme0n1"):
+    conn.execute(
+        "INSERT INTO samples (ts, device, data_units_written, data_units_read, "
+        "percentage_used, bytes_written, bytes_read, power_on_hours) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?)",
+        (ts, device, 1000000, 500000, pu, 512000000000, 256000000000, 8765),
+    )
+    conn.commit()
+
+
+def _open_period(conn, start="2026-09-01T00:00:00+00:00"):
+    ensure_period_open(conn, datetime.fromisoformat(start))
+
+
+# ---------------------------------------------------------------------------
+# Issue #79: Fenris identity and Drive health grouping
+# ---------------------------------------------------------------------------
+
+class TestFenrisIdentity:
+    """TPH-1: Titlebox shows Fenris identity with wolf fallback."""
+
+    @pytest.mark.asyncio
+    async def test_titlebox_shows_wolf_identity(self, tmp_path):
+        """Top titlebox reads 🐺 Fenris by Bongbetic."""
+        app = FenrisTuiApp(
+            store_path=tmp_path / "nonexistent.db",
+            refresh_interval_s=0.2,
+        )
+        async with app.run_test(size=(120, 24)) as pilot:
+            headline = str(app.query_one("#headline-band").render())
+            assert "🐺 Fenris by Bongbetic" in headline
+
+    @pytest.mark.asyncio
+    async def test_titlebox_fallback_without_wolf(self, tmp_path):
+        """Fallback to 'Fenris by Bongbetic' when wolf is unsupported."""
+        app = FenrisTuiApp(
+            store_path=tmp_path / "nonexistent.db",
+            refresh_interval_s=0.2,
+        )
+        # Simulate narrow terminal that can't render wolf
+        async with app.run_test(size=(60, 24)) as pilot:
+            headline = str(app.query_one("#headline-band").render())
+            # Either shows wolf or fallback - both are acceptable
+            assert "Fenris by Bongbetic" in headline
+
+    @pytest.mark.asyncio
+    async def test_wolf_never_shows_tofu(self, tmp_path):
+        """Wolf glyph is never rendered as tofu (unsupported character)."""
+        app = FenrisTuiApp(
+            store_path=tmp_path / "nonexistent.db",
+            refresh_interval_s=0.2,
+        )
+        async with app.run_test(size=(120, 24)) as pilot:
+            headline = str(app.query_one("#headline-band").render())
+            # No replacement character (U+FFFD) should appear
+            assert "\ufffd" not in headline.lower()
+            assert "?" not in headline or "Fenris" in headline
+
+    @pytest.mark.asyncio
+    async def test_lifespan_headline_is_data_surface(self, tmp_path):
+        """Lifespan headline remains a data surface, not app title."""
+        conn = init_store(tmp_path / "test.db")
+        _insert_baseline(conn, tbw_tb=1.0, verified=True)
+        _insert_segment(conn)
+        _open_period(conn)
+        for i in range(30):
+            d = (datetime(2026, 9, 1) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100, coverage=0.95, samples=24)
+        _insert_sample(conn, "2026-09-30T10:00:00+00:00")
+        conn.close()
+
+        app = FenrisTuiApp(store_path=tmp_path / "test.db")
+        async with app.run_test(size=(120, 24)) as pilot:
+            headline = str(app.query_one("#headline-band").render())
+            # Identity appears once, lifespan is separate data
+            assert "🐺 Fenris by Bongbetic" in headline
+            assert "remaining" in headline.lower() or "projection" in headline.lower()
+
+
+class TestSingleMakerCredit:
+    """TPH-1: Single maker-credit placement in title only."""
+
+    @pytest.mark.asyncio
+    async def test_maker_credit_not_in_service_strip(self, tmp_path):
+        """Remove duplicate maker credit from service facts."""
+        conn = init_store(tmp_path / "test.db")
+        _insert_segment(conn)
+        _open_period(conn)
+        for i in range(20):
+            d = (datetime(2026, 9, 10) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100)
+        _insert_sample(conn, "2026-09-30T10:00:00+00:00")
+        conn.close()
+
+        app = FenrisTuiApp(store_path=tmp_path / "test.db")
+        async with app.run_test(size=(120, 24)) as pilot:
+            strip = str(app.query_one("#service-strip").render())
+            # "by Bongbetic" should NOT appear in service strip
+            assert "by Bongbetic" not in strip
+
+    @pytest.mark.asyncio
+    async def test_maker_credit_in_titlebox(self, tmp_path):
+        """Maker credit appears in the titlebox identity."""
+        app = FenrisTuiApp(
+            store_path=tmp_path / "nonexistent.db",
+            refresh_interval_s=0.2,
+        )
+        async with app.run_test(size=(120, 24)) as pilot:
+            headline = str(app.query_one("#headline-band").render())
+            assert "Fenris by Bongbetic" in headline
+
+    @pytest.mark.asyncio
+    async def test_empty_store_no_maker_credit_in_strip(self, tmp_path):
+        """Empty store: no maker credit in service strip."""
+        app = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+        async with app.run_test(size=(120, 24)) as pilot:
+            strip = str(app.query_one("#service-strip").render())
+            assert "by Bongbetic" not in strip
+
+    @pytest.mark.asyncio
+    async def test_store_fault_no_maker_credit_in_strip(self, tmp_path):
+        """Store fault: no maker credit in service strip."""
+        # Create a corrupt store
+        db = tmp_path / "test.db"
+        db.write_bytes(b"not a database")
+        app = FenrisTuiApp(store_path=db)
+        async with app.run_test(size=(120, 24)) as pilot:
+            strip = str(app.query_one("#service-strip").render())
+            assert "by Bongbetic" not in strip
+
+
+class TestDriveHealthVendorWear:
+    """TPH-1: Vendor wear renders under Drive health context."""
+
+    @pytest.mark.asyncio
+    async def test_vendor_wear_in_drive_health(self, tmp_path):
+        """Vendor wear shows with health context, not as Settings."""
+        conn = init_store(tmp_path / "test.db")
+        _insert_segment(conn)
+        _open_period(conn)
+        for i in range(20):
+            d = (datetime(2026, 9, 10) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100)
+        _insert_sample(conn, "2026-09-30T10:00:00+00:00", pu=10)
+        conn.close()
+
+        app = FenrisTuiApp(store_path=tmp_path / "test.db")
+        async with app.run_test(size=(120, 24)) as pilot:
+            health = str(app.query_one("#drive-health").render())
+            # Vendor wear should be present with health context
+            assert "vendor wear" in health.lower()
+            # Should show thermal, spare, errors, etc.
+            assert "temperature" in health.lower()
+            assert "spare" in health.lower()
+            assert "media errors" in health.lower()
+            # Settings section should NOT be a separate heading
+            assert "[bold]Settings[/bold]" not in health
+
+    @pytest.mark.asyncio
+    async def test_missing_wear_values_honest(self, tmp_path):
+        """Missing values remain honest unavailable facts."""
+        conn = init_store(tmp_path / "test.db")
+        _insert_segment(conn)
+        _open_period(conn)
+        for i in range(20):
+            d = (datetime(2026, 9, 10) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100)
+        # Insert sample with zero wear values
+        conn.execute(
+            "INSERT INTO samples (ts, device, data_units_written, data_units_read, "
+            "percentage_used, bytes_written, bytes_read, power_on_hours) "
+            "VALUES (?, ?, ?, ?, ?, ?, ?, ?)",
+            ("2026-09-30T10:00:00+00:00", "/dev/nvme0n1", 0, 0, 0, 0, 0, 0),
+        )
+        conn.commit()
+        conn.close()
+
+        app = FenrisTuiApp(store_path=tmp_path / "test.db")
+        async with app.run_test(size=(120, 24)) as pilot:
+            health = str(app.query_one("#drive-health").render())
+            # Should show 0% used or honest zero, not crash
+            assert "vendor wear" in health.lower()
+
+
+class TestPreservedBehavior:
+    """Preserve existing continuity, pause, quit, auth behavior."""
+
+    @pytest.mark.asyncio
+    async def test_continuity_preserved(self, tmp_path):
+        """Continuity wording preserved in service strip."""
+        conn = init_store(tmp_path / "test.db")
+        _open_period(conn)
+        conn.close()
+        app = FenrisTuiApp(store_path=tmp_path / "test.db")
+
+        with patch("fenris.tui.query_service_state", return_value={
+            "boot_enabled": True, "timer_active": True,
+            "last_collect_ok": True, "last_collect_age_s": 60,
+            "last_collect_reason": None,
+        }):
+            async with app.run_test(size=(120, 24)) as pilot:
+                strip = str(app.query_one("#service-strip").render())
+                assert "continuity" in strip.lower()
+                assert "monitoring" in strip.lower()
+
+    @pytest.mark.asyncio
+    async def test_quit_rail_preserved(self, tmp_path):
+        """Separate q QUIT TUI rail preserved."""
+        app = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+        async with app.run_test(size=(120, 24)) as pilot:
+            rail = str(app.query_one("#quit-rail").render())
+            assert "q QUIT TUI" in rail
+
+    @pytest.mark.asyncio
+    async def test_auth_banner_preserved(self, tmp_path):
+        """Polkit authentication banner lifecycle preserved."""
+        app = FenrisTuiApp(
+            store_path=tmp_path / "nonexistent.db",
+            refresh_interval_s=0.2,
+        )
+        async with app.run_test(size=(120, 24)) as pilot:
+            headline = str(app.query_one("#headline-band").render())
+            assert "polkit" in headline.lower()
+            # Should clear after first tick
+            await pilot.pause(0.25)
+            headline_after = str(app.query_one("#headline-band").render())
+            assert "polkit" not in headline_after.lower()
+
+    @pytest.mark.asyncio
+    async def test_deliberate_pause_block_preserved(self, tmp_path):
+        """Deliberate pause banner preserved."""
+        db = tmp_path / "test.db"
+        conn = init_store(db)
+        conn.execute(
+            "INSERT INTO monitoring_periods (started_at, ended_at, end_cause) "
+            "VALUES (?, ?, ?)",
+            ("2026-09-01T09:00:00+00:00", "2026-09-01T10:00:00+00:00", "user_disabled"),
+        )
+        conn.commit()
+        conn.close()
+        app = FenrisTuiApp(store_path=db)
+
+        with patch("fenris.tui.query_service_state", return_value={
+            "boot_enabled": False, "timer_active": False,
+            "last_collect_ok": None, "last_collect_age_s": None,
+            "last_collect_reason": None,
+        }):
+            async with app.run_test(size=(120, 24)) as pilot:
+                banner = str(app.query_one("#paused-banner").render())
+                assert "paused" in banner.lower()
+                assert "deliberate" in banner.lower()

tests/test_issue_80.py

@@ -0,0 +1,482 @@
+"""Integration tests for issue #80: Persist accessible colour and motion preferences.
+
+Covers:
+  - AC80-1: Amber/Nord/High Contrast presets with Amber default
+  - AC80-2: t preset and m motion controls with clickable equivalents
+  - AC80-3: Persistent user-scoped XDG TUI preferences
+  - AC80-4: Reduced motion makes Monitoring steady
+  - AC80-5: Theme roles for graph rendering
+  - AC80-6: Headless interaction tests with temporary user config
+"""
+import json
+import os
+from datetime import datetime, timedelta, timezone
+from pathlib import Path
+from unittest.mock import patch
+
+import pytest
+import sys
+
+sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
+
+from fenris.store import init_store
+from fenris.preferences import (
+    load_preferences,
+    save_preferences,
+    get_preference_path,
+)
+from fenris.themes import get_theme, get_graph_colors, THEME_NAMES
+from fenris.status_composition import (
+    StatusState,
+    compose_status,
+    render_status_tui,
+)
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+def _clock(year=2026, month=9, day=30, hour=12):
+    return datetime(year, month, day, hour, 0, 0, tzinfo=timezone.utc)
+
+
+def _insert_segment(conn, opened_at="2026-09-01T00:00:00+00:00",
+                     identity_key="nqn.test", degraded=False,
+                     mn="Samsung SSD 970 EVO Plus 1TB"):
+    conn.execute(
+        "INSERT INTO controller_segments "
+        "(opened_at, identity_key, identity_degraded, subnqn, sn, mn, fr, vid, ssvid, transport) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)",
+        (opened_at, identity_key, degraded, "nqn.test", "SN123", mn, "FW1",
+         "0x144d", "0x144d", "pcie"),
+    )
+    conn.commit()
+
+
+def _insert_day(conn, day, bw=1024*1024*100, coverage=0.95, samples=24):
+    conn.execute(
+        "INSERT INTO day_aggregates (day, active_seconds, idle_seconds, powered_off_seconds, "
+        "unknown_seconds, bytes_written_delta, bytes_read_delta, sample_count, coverage) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)",
+        (day, 3600, 0, 0, 0, bw, 0, samples, coverage),
+    )
+    conn.commit()
+
+
+def _insert_sample(conn, ts, device="/dev/nvme0n1"):
+    conn.execute(
+        "INSERT INTO samples (ts, device, data_units_written, data_units_read, "
+        "percentage_used, bytes_written, bytes_read, power_on_hours) "
+        "VALUES (?, ?, ?, ?, ?, ?, ?, ?)",
+        (ts, device, 1000000, 500000, 5, 512000000000, 256000000000, 8765),
+    )
+    conn.commit()
+
+
+def _open_period(conn, start="2026-09-01T00:00:00+00:00"):
+    from fenris.monitoring_periods import ensure_period_open
+    ensure_period_open(conn, datetime.fromisoformat(start))
+
+
+def _make_prefs_dir(tmp_path: Path) -> Path:
+    config_home = tmp_path / ".config"
+    config_home.mkdir(parents=True, exist_ok=True)
+    return config_home
+
+
+# ---------------------------------------------------------------------------
+# AC80-1: Preset loading and application
+# ---------------------------------------------------------------------------
+
+class TestPresetLoading:
+    """Themes load from preferences and apply to the TUI."""
+
+    @pytest.mark.asyncio
+    async def test_tui_applies_amber_theme_by_default(self, tmp_path):
+        """TUI starts with the Amber theme when no preferences exist."""
+        from fenris.tui import FenrisTuiApp
+
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            app = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            async with app.run_test() as pilot:
+                # Theme should be fenris-amber
+                assert app.theme == "fenris-amber"
+
+    @pytest.mark.asyncio
+    async def test_tui_applies_nord_theme_from_prefs(self, tmp_path):
+        """TUI applies Nord theme from saved preferences."""
+        from fenris.tui import FenrisTuiApp
+
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            save_preferences(theme="nord", reduced_motion=False)
+            app = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            async with app.run_test() as pilot:
+                assert app.theme == "fenris-nord"
+
+    @pytest.mark.asyncio
+    async def test_tui_applies_high_contrast_from_prefs(self, tmp_path):
+        """TUI applies High Contrast theme from saved preferences."""
+        from fenris.tui import FenrisTuiApp
+
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            save_preferences(theme="high_contrast", reduced_motion=False)
+            app = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            async with app.run_test() as pilot:
+                assert app.theme == "fenris-high-contrast"
+
+    @pytest.mark.asyncio
+    async def test_tui_applies_reduced_motion_from_prefs(self, tmp_path):
+        """TUI respects reduced_motion preference."""
+        from fenris.tui import FenrisTuiApp
+        from fenris.status_composition import compose_status
+
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            save_preferences(theme="amber", reduced_motion=True)
+            app = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            async with app.run_test() as pilot:
+                # The app should have reduced_motion set
+                assert app._reduced_motion is True
+
+
+# ---------------------------------------------------------------------------
+# AC80-2: t and m key bindings
+# ---------------------------------------------------------------------------
+
+class TestKeyBindings:
+    """t cycles presets and m toggles reduced motion."""
+
+    @pytest.mark.asyncio
+    async def test_t_binding_exists(self, tmp_path):
+        """The TUI has a 't' binding for theme cycling."""
+        from fenris.tui import FenrisTuiApp
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            app = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            binding_keys = {b.key for b in app.BINDINGS}
+            assert "t" in binding_keys
+
+    @pytest.mark.asyncio
+    async def test_m_binding_exists(self, tmp_path):
+        """The TUI has an 'm' binding for motion toggle."""
+        from fenris.tui import FenrisTuiApp
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            app = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            binding_keys = {b.key for b in app.BINDINGS}
+            assert "m" in binding_keys
+
+    @pytest.mark.asyncio
+    async def test_t_cycles_through_themes(self, tmp_path):
+        """Pressing t cycles through the three presets."""
+        from fenris.tui import FenrisTuiApp
+
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            app = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            async with app.run_test() as pilot:
+                # Start at amber
+                assert app.theme == "fenris-amber"
+
+                # Press t to cycle
+                await pilot.press("t")
+                await pilot.pause()
+
+                # Should be nord or high_contrast now
+                assert app.theme in ("fenris-nord", "fenris-high-contrast")
+
+    @pytest.mark.asyncio
+    async def test_m_toggles_reduced_motion(self, tmp_path):
+        """Pressing m toggles reduced motion."""
+        from fenris.tui import FenrisTuiApp
+
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            app = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            async with app.run_test() as pilot:
+                # Start with normal motion
+                assert app._reduced_motion is False
+
+                # Press m to toggle
+                await pilot.press("m")
+                await pilot.pause()
+
+                # Should be reduced motion now
+                assert app._reduced_motion is True
+
+                # Press m again to toggle back
+                await pilot.press("m")
+                await pilot.pause()
+
+                assert app._reduced_motion is False
+
+
+# ---------------------------------------------------------------------------
+# AC80-3: Persistence across restart
+# ---------------------------------------------------------------------------
+
+class TestPersistence:
+    """Preferences survive app restart."""
+
+    @pytest.mark.asyncio
+    async def test_theme_survives_restart(self, tmp_path):
+        """Theme preference persists across TUI restart."""
+        from fenris.tui import FenrisTuiApp
+
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            # First run: change theme
+            app1 = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            async with app1.run_test() as pilot:
+                await pilot.press("t")
+                await pilot.pause()
+                theme_after_t = app1.theme
+                assert theme_after_t != "fenris-amber"
+
+            # Second run: theme should persist
+            app2 = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            async with app2.run_test() as pilot:
+                assert app2.theme == theme_after_t
+
+    @pytest.mark.asyncio
+    async def test_reduced_motion_survives_restart(self, tmp_path):
+        """Reduced motion preference persists across TUI restart."""
+        from fenris.tui import FenrisTuiApp
+
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            # First run: toggle motion
+            app1 = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            async with app1.run_test() as pilot:
+                await pilot.press("m")
+                await pilot.pause()
+                assert app1._reduced_motion is True
+
+            # Second run: motion should persist
+            app2 = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            async with app2.run_test() as pilot:
+                assert app2._reduced_motion is True
+
+
+# ---------------------------------------------------------------------------
+# AC80-4: Reduced motion makes Monitoring steady
+# ---------------------------------------------------------------------------
+
+class TestReducedMotion:
+    """Reduced motion disables the Monitoring dot blink."""
+
+    def test_reduced_motion_disables_blink(self, tmp_path):
+        """compose_status with reduced_motion=True → should_blink=False."""
+        conn = init_store(tmp_path / "test.db")
+        _insert_segment(conn)
+        _open_period(conn)
+        now = _clock()
+        ts = (now - timedelta(minutes=2)).isoformat()
+        _insert_sample(conn, ts)
+        for i in range(20):
+            d = (datetime(2026, 9, 10) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100)
+        conn.close()
+
+        conn = init_store(tmp_path / "test.db")
+        svc = {
+            "boot_enabled": True, "timer_active": True,
+            "last_collect_ok": True, "last_collect_age_s": 120,
+            "last_collect_reason": None,
+        }
+        comp = compose_status(conn, svc, now, store_fault=None, newer_schema=None,
+                              reduced_motion=True)
+        assert comp.state == StatusState.MONITORING
+        assert comp.should_blink is False
+        conn.close()
+
+    def test_normal_motion_allows_blink(self, tmp_path):
+        """compose_status with reduced_motion=False → should_blink=True for Monitoring."""
+        conn = init_store(tmp_path / "test.db")
+        _insert_segment(conn)
+        _open_period(conn)
+        now = _clock()
+        ts = (now - timedelta(minutes=2)).isoformat()
+        _insert_sample(conn, ts)
+        for i in range(20):
+            d = (datetime(2026, 9, 10) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100)
+        conn.close()
+
+        conn = init_store(tmp_path / "test.db")
+        svc = {
+            "boot_enabled": True, "timer_active": True,
+            "last_collect_ok": True, "last_collect_age_s": 120,
+            "last_collect_reason": None,
+        }
+        comp = compose_status(conn, svc, now, store_fault=None, newer_schema=None,
+                              reduced_motion=False)
+        assert comp.state == StatusState.MONITORING
+        assert comp.should_blink is True
+        conn.close()
+
+    @pytest.mark.asyncio
+    async def test_framework_reduced_motion_honoured(self, tmp_path):
+        """TUI honours Textual's reduced_motion signal."""
+        from fenris.tui import FenrisTuiApp
+
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            app = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            async with app.run_test() as pilot:
+                # The app should check for reduced motion on mount
+                assert hasattr(app, '_reduced_motion')
+
+
+# ---------------------------------------------------------------------------
+# AC80-5: Theme roles for graph rendering
+# ---------------------------------------------------------------------------
+
+class TestGraphThemeRoles:
+    """Graph uses theme-derived colours for each bar role."""
+
+    def test_graph_colors_available_for_all_themes(self):
+        """Every theme provides all required graph colour roles."""
+        required_roles = {"allocated", "unallocated", "gap", "zero", "partial"}
+        for name in THEME_NAMES:
+            colors = get_graph_colors(name)
+            assert required_roles.issubset(set(colors.keys())),                 "Theme %s missing roles: %s" % (name, required_roles - set(colors.keys()))
+
+    def test_graph_colors_are_strings(self):
+        """All graph colour values are strings (Textual CSS colour values)."""
+        for name in THEME_NAMES:
+            colors = get_graph_colors(name)
+            for role, color in colors.items():
+                assert isinstance(color, str),                     "Theme %s role %s has non-string color: %s" % (name, role, color)
+
+
+# ---------------------------------------------------------------------------
+# AC80-6: Safe persistence — failure modes
+# ---------------------------------------------------------------------------
+
+class TestSafePersistence:
+    """Preference failures never crash the dashboard."""
+
+    @pytest.mark.asyncio
+    async def test_corrupt_prefs_does_not_crash_tui(self, tmp_path):
+        """Corrupt preference file does not prevent TUI from starting."""
+        from fenris.tui import FenrisTuiApp
+
+        config_home = _make_prefs_dir(tmp_path)
+        fenris_dir = config_home / "fenris"
+        fenris_dir.mkdir(parents=True, exist_ok=True)
+        (fenris_dir / "preferences.json").write_text("{bad json!!!")
+
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            app = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            async with app.run_test() as pilot:
+                # TUI should start with default theme
+                assert app.theme == "fenris-amber"
+                # Dashboard should be functional
+                headline = str(app.query_one("#headline-band").render())
+                assert headline is not None
+
+    @pytest.mark.asyncio
+    async def test_readonly_config_dir_does_not_crash(self, tmp_path):
+        """Read-only config directory does not prevent TUI from starting."""
+        from fenris.tui import FenrisTuiApp
+
+        config_home = _make_prefs_dir(tmp_path)
+        fenris_dir = config_home / "fenris"
+        fenris_dir.mkdir(parents=True, exist_ok=True)
+        fenris_dir.chmod(0o555)
+
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            app = FenrisTuiApp(store_path=tmp_path / "nonexistent.db")
+            async with app.run_test() as pilot:
+                # TUI should start without crash
+                assert app.theme in ("fenris-amber", "fenris-nord", "fenris-high-contrast")
+
+
+# ---------------------------------------------------------------------------
+# CLI isolation (preferences do not affect CLI)
+# ---------------------------------------------------------------------------
+
+class TestCLIIsolation:
+    """Preferences are TUI-only — CLI status is independent."""
+
+    def test_cli_status_unchanged_by_theme(self, tmp_path):
+        """fenris status output does not change based on theme preference."""
+        from fenris.status import get_status
+
+        conn = init_store(tmp_path / "test.db")
+        _insert_segment(conn)
+        _open_period(conn)
+        now = _clock()
+        ts = (now - timedelta(minutes=2)).isoformat()
+        _insert_sample(conn, ts)
+        for i in range(20):
+            d = (datetime(2026, 9, 10) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100)
+        conn.close()
+
+        config_home = _make_prefs_dir(tmp_path)
+
+        # With amber theme
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            save_preferences(theme="amber", reduced_motion=False)
+            status_amber = get_status(
+                store_path=tmp_path / "test.db",
+                clock_now=now,
+                query_services=False,
+                query_journal=False,
+            )
+
+        # With high contrast theme
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            save_preferences(theme="high_contrast", reduced_motion=True)
+            status_hc = get_status(
+                store_path=tmp_path / "test.db",
+                clock_now=now,
+                query_services=False,
+                query_journal=False,
+            )
+
+        # Status output should be identical
+        assert status_amber == status_hc
+
+    def test_cli_status_unchanged_by_motion(self, tmp_path):
+        """fenris status output does not change based on reduced_motion preference."""
+        from fenris.status import get_status
+
+        conn = init_store(tmp_path / "test.db")
+        _insert_segment(conn)
+        _open_period(conn)
+        now = _clock()
+        ts = (now - timedelta(minutes=2)).isoformat()
+        _insert_sample(conn, ts)
+        for i in range(20):
+            d = (datetime(2026, 9, 10) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(conn, d, bw=1024*1024*100)
+        conn.close()
+
+        config_home = _make_prefs_dir(tmp_path)
+
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            save_preferences(theme="amber", reduced_motion=False)
+            status_normal = get_status(
+                store_path=tmp_path / "test.db",
+                clock_now=now,
+                query_services=False,
+                query_journal=False,
+            )
+
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            save_preferences(theme="amber", reduced_motion=True)
+            status_reduced = get_status(
+                store_path=tmp_path / "test.db",
+                clock_now=now,
+                query_services=False,
+                query_journal=False,
+            )
+
+        assert status_normal == status_reduced

tests/test_monitor.py

@@ -0,0 +1,325 @@
+"""Tests for fenris-monitor helper.
+
+Spec: §8.4, §8.5, §8.6, §8.7
+"""
+import json
+import sqlite3
+from datetime import datetime, timezone
+from pathlib import Path
+from unittest.mock import patch, MagicMock
+
+import pytest
+
+import sys
+sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
+
+from fenris.monitor import (
+    cmd_enable,
+    cmd_disable,
+    cmd_collect,
+    cmd_baseline_set,
+    cmd_baseline_clear,
+    is_root,
+)
+from fenris.store import init_store
+
+
+@pytest.fixture
+def tmp_store(tmp_path):
+    """Create a temporary observation store."""
+    store_path = tmp_path / "observations.db"
+    conn = init_store(store_path)
+    yield conn
+    conn.close()
+
+
+@pytest.fixture
+def store_path(tmp_path):
+    """Return path to a temporary observation store."""
+    return tmp_path / "observations.db"
+
+
+class TestIsRoot:
+    def test_root_returns_true(self):
+        with patch("os.geteuid", return_value=0):
+            assert is_root() is True
+
+    def test_non_root_returns_false(self):
+        with patch("os.geteuid", return_value=1000):
+            assert is_root() is False
+
+
+class TestEnableIdempotentMatrix:
+    """§8.6: Period-row idempotent matrix."""
+
+    def test_first_enable_creates_missing_store(self, store_path):
+        """A fresh package install has a store directory but no database yet."""
+        args = MagicMock(now=False, store_path=store_path)
+
+        with patch("fenris.monitor.subprocess") as mock_sub:
+            mock_sub.run.return_value = MagicMock(returncode=0)
+            cmd_enable(args)
+
+        conn = init_store(store_path)
+        row = conn.execute(
+            "SELECT ended_at FROM monitoring_periods WHERE ended_at IS NULL"
+        ).fetchone()
+        conn.close()
+        assert row is not None
+
+    def test_first_opens_period(self, store_path):
+        """First-ever enable opens a period at the enable moment."""
+        # Initialize store
+        init_store(store_path)
+        
+        args = MagicMock(now=False, store_path=store_path)
+
+        with patch("fenris.monitor.subprocess") as mock_sub:
+            mock_sub.run.return_value = MagicMock(returncode=0)
+            cmd_enable(args)
+
+        # Period should be open
+        conn = init_store(store_path)
+        cursor = conn.execute(
+            "SELECT ended_at FROM monitoring_periods WHERE ended_at IS NULL"
+        )
+        assert cursor.fetchone() is not None
+        conn.close()
+
+    def test_resume_with_open_period_noop(self, store_path):
+        """Resume with open period: no-op (gap stays inside as unknown)."""
+        # Initialize store and open a period
+        conn = init_store(store_path)
+        now = datetime.now(timezone.utc)
+        conn.execute(
+            "INSERT INTO monitoring_periods (started_at) VALUES (?)",
+            (now.isoformat(),),
+        )
+        conn.commit()
+        conn.close()
+
+        args = MagicMock(now=True, store_path=store_path)
+
+        with patch("fenris.monitor.subprocess") as mock_sub:
+            mock_sub.run.return_value = MagicMock(returncode=0)
+            cmd_enable(args)
+
+        # Should still have exactly one open period
+        conn = init_store(store_path)
+        cursor = conn.execute(
+            "SELECT COUNT(*) FROM monitoring_periods WHERE ended_at IS NULL"
+        )
+        assert cursor.fetchone()[0] == 1
+        conn.close()
+
+    def test_resume_with_no_period_opens_new(self, store_path):
+        """Resume with no open period opens a new row."""
+        # Initialize store and close any existing period
+        conn = init_store(store_path)
+        conn.execute(
+            "UPDATE monitoring_periods SET ended_at = ?, end_cause = ?",
+            (datetime.now(timezone.utc).isoformat(), "user_disabled"),
+        )
+        conn.commit()
+        conn.close()
+
+        args = MagicMock(now=True, store_path=store_path)
+
+        with patch("fenris.monitor.subprocess") as mock_sub:
+            mock_sub.run.return_value = MagicMock(returncode=0)
+            cmd_enable(args)
+
+        # Should have a new open period
+        conn = init_store(store_path)
+        cursor = conn.execute(
+            "SELECT COUNT(*) FROM monitoring_periods WHERE ended_at IS NULL"
+        )
+        assert cursor.fetchone()[0] == 1
+        conn.close()
+
+
+class TestDisableIdempotentMatrix:
+    """§8.6: Period-row idempotent matrix."""
+
+    def test_pause_with_open_period_closes_user_disabled(self, store_path):
+        """Pause with open period closes it user_disabled."""
+        # Initialize store and open a period
+        conn = init_store(store_path)
+        now = datetime.now(timezone.utc)
+        conn.execute(
+            "INSERT INTO monitoring_periods (started_at) VALUES (?)",
+            (now.isoformat(),),
+        )
+        conn.commit()
+        conn.close()
+
+        args = MagicMock(now=True, store_path=store_path)
+
+        with patch("fenris.monitor.subprocess") as mock_sub:
+            mock_sub.run.return_value = MagicMock(returncode=0)
+            cmd_disable(args)
+
+        # Period should be closed with user_disabled
+        conn = init_store(store_path)
+        cursor = conn.execute(
+            "SELECT end_cause FROM monitoring_periods WHERE ended_at IS NOT NULL"
+        )
+        assert cursor.fetchone()[0] == "user_disabled"
+        conn.close()
+
+    def test_pause_without_open_period_noop(self, store_path):
+        """Pause otherwise no-ops."""
+        # Initialize store
+        init_store(store_path)
+
+        args = MagicMock(now=True, store_path=store_path)
+
+        with patch("fenris.monitor.subprocess") as mock_sub:
+            mock_sub.run.return_value = MagicMock(returncode=0)
+            cmd_disable(args)
+
+        # No periods should exist
+        conn = init_store(store_path)
+        cursor = conn.execute("SELECT COUNT(*) FROM monitoring_periods")
+        assert cursor.fetchone()[0] == 0
+        conn.close()
+
+    def test_raw_systemctl_stop_never_records_user_disabled(self, store_path):
+        """Raw systemctl stop outside helper never records user_disabled."""
+        # Initialize store and open a period
+        conn = init_store(store_path)
+        now = datetime.now(timezone.utc)
+        conn.execute(
+            "INSERT INTO monitoring_periods (started_at) VALUES (?)",
+            (now.isoformat(),),
+        )
+        conn.commit()
+
+        # Simulate raw systemctl stop (no monitor involved)
+        # The period stays open - only the sanctioned path closes it
+        cursor = conn.execute(
+            "SELECT end_cause FROM monitoring_periods WHERE ended_at IS NULL"
+        )
+        assert cursor.fetchone() is not None  # Still open
+        conn.close()
+
+
+class TestCollectTrigger:
+    """§8.7: On-demand collection via helper path."""
+
+    def test_collect_triggers_systemctl_start(self):
+        """Collect starts fenris-collect.service synchronously."""
+        args = MagicMock()
+
+        with patch("fenris.monitor.subprocess") as mock_sub:
+            mock_sub.run.return_value = MagicMock(returncode=0)
+            cmd_collect(args)
+
+        mock_sub.run.assert_called_once_with(
+            ["systemctl", "start", "fenris-collect.service"],
+            capture_output=True,
+            text=True,
+        )
+
+    def test_collect_failure_exits_nonzero(self):
+        """Collect failure exits with nonzero status."""
+        args = MagicMock()
+
+        with patch("fenris.monitor.subprocess") as mock_sub:
+            mock_sub.run.return_value = MagicMock(
+                returncode=1, stderr="Unit not found"
+            )
+            with pytest.raises(SystemExit) as exc_info:
+                cmd_collect(args)
+            assert exc_info.value.code == 1
+
+
+class TestBaselinePersistence:
+    """PR-14: Baseline persistence behind polkit-guarded helper."""
+
+    def test_baseline_set_persists(self, store_path):
+        """baseline set persists the baseline row."""
+        # Initialize store
+        init_store(store_path)
+
+        args = MagicMock(
+            store_path=store_path,
+            baseline_json=json.dumps(
+                {
+                    "tbw_terabytes": 600,
+                    "source_url": "https://example.com/spec",
+                    "document_revision": "rev1",
+                    "entry_date": "2024-01-01",
+                    "model_string": "Samsung 990 Pro",
+                    "nominal_capacity_bytes": 2000000000000,
+                }
+            )
+        )
+
+        cmd_baseline_set(args)
+
+        conn = init_store(store_path)
+        cursor = conn.execute("SELECT * FROM endurance_baseline")
+        row = cursor.fetchone()
+        assert row is not None
+        assert row[1] == 600.0  # tbw_terabytes
+        conn.close()
+
+    def test_baseline_set_replaces_existing(self, store_path):
+        """baseline set replaces any existing baseline."""
+        # Initialize store and insert initial baseline
+        conn = init_store(store_path)
+        now = datetime.now(timezone.utc)
+        conn.execute(
+            "INSERT INTO endurance_baseline (tbw_terabytes, source_url, "
+            "document_revision, entry_date, model_string, nominal_capacity_bytes, "
+            "created_at, updated_at) VALUES (?, ?, ?, ?, ?, ?, ?, ?)",
+            (400, "old", "v1", "2023-01-01", "Old Model", 1000000000000,
+             now.isoformat(), now.isoformat()),
+        )
+        conn.commit()
+        conn.close()
+
+        args = MagicMock(
+            store_path=store_path,
+            baseline_json=json.dumps(
+                {
+                    "tbw_terabytes": 600,
+                    "source_url": "new",
+                    "document_revision": "v2",
+                    "entry_date": "2024-01-01",
+                    "model_string": "New Model",
+                    "nominal_capacity_bytes": 2000000000000,
+                }
+            )
+        )
+
+        cmd_baseline_set(args)
+
+        conn = init_store(store_path)
+        cursor = conn.execute("SELECT COUNT(*) FROM endurance_baseline")
+        assert cursor.fetchone()[0] == 1  # Only one row
+        conn.close()
+
+    def test_baseline_clear_removes(self, store_path):
+        """baseline clear removes the baseline."""
+        # Initialize store and insert baseline
+        conn = init_store(store_path)
+        now = datetime.now(timezone.utc)
+        conn.execute(
+            "INSERT INTO endurance_baseline (tbw_terabytes, source_url, "
+            "document_revision, entry_date, model_string, nominal_capacity_bytes, "
+            "created_at, updated_at) VALUES (?, ?, ?, ?, ?, ?, ?, ?)",
+            (400, "src", "v1", "2024-01-01", "Model", 1000000000000,
+             now.isoformat(), now.isoformat()),
+        )
+        conn.commit()
+        conn.close()
+
+        args = MagicMock(store_path=store_path)
+        cmd_baseline_clear(args)
+
+        conn = init_store(store_path)
+        cursor = conn.execute("SELECT COUNT(*) FROM endurance_baseline")
+        assert cursor.fetchone()[0] == 0
+        conn.close()

tests/test_preferences.py

@@ -0,0 +1,301 @@
+"""Tests for user-scoped TUI preferences (issue #80).
+
+Covers:
+  - Preference load/save with safe defaults
+  - XDG_CONFIG_HOME user-scoped persistence
+  - Amber default theme, normal-motion default
+  - Invalid/unreadable/unwritable preference data does not crash
+  - Theme presets: Amber, Nord, High Contrast
+  - Reduced motion preference persistence
+  - CLI status and collector behaviour unchanged by preferences
+"""
+import json
+import os
+from pathlib import Path
+from unittest.mock import patch
+
+import pytest
+import sys
+
+sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
+
+from fenris.preferences import (
+    load_preferences,
+    save_preferences,
+    get_preference_path,
+    PREFERENCE_FILE_NAME,
+    VALID_THEMES,
+    DEFAULT_THEME,
+    DEFAULT_REDUCED_MOTION,
+)
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+def _make_prefs_dir(tmp_path: Path) -> Path:
+    """Create a fake XDG_CONFIG_HOME with fenris subdir."""
+    config_home = tmp_path / ".config"
+    config_home.mkdir(parents=True, exist_ok=True)
+    return config_home
+
+
+# ---------------------------------------------------------------------------
+# Preference path tests
+# ---------------------------------------------------------------------------
+
+class TestPreferencePath:
+    """Preference file lives at XDG_CONFIG_HOME/fenris/preferences.json."""
+
+    def test_uses_xdg_config_home(self, tmp_path):
+        """Path respects XDG_CONFIG_HOME environment variable."""
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            path = get_preference_path()
+            assert path == config_home / "fenris" / PREFERENCE_FILE_NAME
+
+    def test_default_path_fallback(self):
+        """When XDG_CONFIG_HOME is unset, falls back to ~/.config."""
+        with patch.dict(os.environ, {}, clear=True):
+            # Remove XDG_CONFIG_HOME if present
+            os.environ.pop("XDG_CONFIG_HOME", None)
+            path = get_preference_path()
+            assert "fenris" in str(path)
+            assert PREFERENCE_FILE_NAME in str(path)
+
+
+# ---------------------------------------------------------------------------
+# Default preferences tests
+# ---------------------------------------------------------------------------
+
+class TestDefaults:
+    """When no preference file exists, defaults are returned."""
+
+    def test_default_theme_is_amber(self):
+        """Amber is the default theme preset."""
+        assert DEFAULT_THEME == "amber"
+
+    def test_default_reduced_motion_is_false(self):
+        """Normal motion is the default."""
+        assert DEFAULT_REDUCED_MOTION is False
+
+    def test_valid_themes_are_all_presets(self):
+        """Three valid presets exist."""
+        assert VALID_THEMES == {"amber", "nord", "high_contrast"}
+
+    def test_load_returns_defaults_when_no_file(self, tmp_path):
+        """Missing preference file returns safe defaults."""
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(tmp_path)}):
+            prefs = load_preferences()
+            assert prefs["theme"] == "amber"
+            assert prefs["reduced_motion"] is False
+
+
+# ---------------------------------------------------------------------------
+# Save and load round-trip tests
+# ---------------------------------------------------------------------------
+
+class TestRoundTrip:
+    """Preferences survive save → load."""
+
+    def test_save_and_load_basic(self, tmp_path):
+        """Basic theme and motion save/load."""
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            save_preferences(theme="nord", reduced_motion=True)
+            prefs = load_preferences()
+            assert prefs["theme"] == "nord"
+            assert prefs["reduced_motion"] is True
+
+    def test_save_creates_directory(self, tmp_path):
+        """Save creates the fenris config directory if missing."""
+        config_home = tmp_path / ".config"
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            save_preferences(theme="amber", reduced_motion=False)
+            assert (config_home / "fenris" / PREFERENCE_FILE_NAME).exists()
+
+    def test_overwrites_existing(self, tmp_path):
+        """Second save overwrites the first."""
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            save_preferences(theme="amber", reduced_motion=False)
+            save_preferences(theme="high_contrast", reduced_motion=True)
+            prefs = load_preferences()
+            assert prefs["theme"] == "high_contrast"
+            assert prefs["reduced_motion"] is True
+
+    def test_all_themes_round_trip(self, tmp_path):
+        """Every valid theme saves and loads correctly."""
+        config_home = _make_prefs_dir(tmp_path)
+        for theme in VALID_THEMES:
+            with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+                save_preferences(theme=theme, reduced_motion=False)
+                prefs = load_preferences()
+                assert prefs["theme"] == theme
+
+
+# ---------------------------------------------------------------------------
+# Safe failure tests — invalid/unreadable/unwritable
+# ---------------------------------------------------------------------------
+
+class TestSafeFailures:
+    """Invalid data never crashes the dashboard."""
+
+    def test_corrupt_json_returns_defaults(self, tmp_path):
+        """Malformed JSON returns safe defaults."""
+        config_home = _make_prefs_dir(tmp_path)
+        fenris_dir = config_home / "fenris"
+        fenris_dir.mkdir(parents=True, exist_ok=True)
+        (fenris_dir / PREFERENCE_FILE_NAME).write_text("{corrupt json!!")
+
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            prefs = load_preferences()
+            assert prefs["theme"] == "amber"
+            assert prefs["reduced_motion"] is False
+
+    def test_unknown_theme_returns_default(self, tmp_path):
+        """Unrecognized theme value falls back to amber."""
+        config_home = _make_prefs_dir(tmp_path)
+        fenris_dir = config_home / "fenris"
+        fenris_dir.mkdir(parents=True, exist_ok=True)
+        (fenris_dir / PREFERENCE_FILE_NAME).write_text(
+            json.dumps({"theme": "neon-pink", "reduced_motion": False})
+        )
+
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            prefs = load_preferences()
+            assert prefs["theme"] == "amber"
+
+    def test_missing_keys_get_defaults(self, tmp_path):
+        """Partial preference file fills in missing keys."""
+        config_home = _make_prefs_dir(tmp_path)
+        fenris_dir = config_home / "fenris"
+        fenris_dir.mkdir(parents=True, exist_ok=True)
+        (fenris_dir / PREFERENCE_FILE_NAME).write_text(
+            json.dumps({"theme": "nord"})
+        )
+
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            prefs = load_preferences()
+            assert prefs["theme"] == "nord"
+            assert prefs["reduced_motion"] is False
+
+    def test_unreadable_file_returns_defaults(self, tmp_path):
+        """Permission denied on preference file returns defaults."""
+        config_home = _make_prefs_dir(tmp_path)
+        fenris_dir = config_home / "fenris"
+        fenris_dir.mkdir(parents=True, exist_ok=True)
+        pref_file = fenris_dir / PREFERENCE_FILE_NAME
+        pref_file.write_text(json.dumps({"theme": "nord"}))
+        pref_file.chmod(0o000)
+
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            prefs = load_preferences()
+            # Should fall back to defaults without crashing
+            assert prefs["theme"] in VALID_THEMES
+
+    def test_unwritable_location_returns_defaults(self, tmp_path):
+        """Read-only config directory returns defaults without crashing."""
+        config_home = _make_prefs_dir(tmp_path)
+        fenris_dir = config_home / "fenris"
+        fenris_dir.mkdir(parents=True, exist_ok=True)
+        fenris_dir.chmod(0o555)
+
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            # Should not raise
+            save_preferences(theme="nord", reduced_motion=True)
+            prefs = load_preferences()
+            # Either saved successfully or fell back — either way, no crash
+            assert prefs["theme"] in VALID_THEMES
+
+    def test_non_json_file_returns_defaults(self, tmp_path):
+        """A file that isn't JSON returns defaults."""
+        config_home = _make_prefs_dir(tmp_path)
+        fenris_dir = config_home / "fenris"
+        fenris_dir.mkdir(parents=True, exist_ok=True)
+        (fenris_dir / PREFERENCE_FILE_NAME).write_text("this is not json")
+
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            prefs = load_preferences()
+            assert prefs["theme"] == "amber"
+
+    def test_wrong_type_for_reduced_motion(self, tmp_path):
+        """Non-boolean reduced_motion falls back to default."""
+        config_home = _make_prefs_dir(tmp_path)
+        fenris_dir = config_home / "fenris"
+        fenris_dir.mkdir(parents=True, exist_ok=True)
+        (fenris_dir / PREFERENCE_FILE_NAME).write_text(
+            json.dumps({"theme": "amber", "reduced_motion": "yes"})
+        )
+
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            prefs = load_preferences()
+            assert prefs["reduced_motion"] is False
+
+
+# ---------------------------------------------------------------------------
+# CLI/collector isolation tests
+# ---------------------------------------------------------------------------
+
+class TestCLIIsolation:
+    """TUI display preferences do not affect CLI status or collector."""
+
+    def test_cli_status_ignores_preferences(self, tmp_path):
+        """fenris status output is independent of TUI preferences."""
+        from fenris.status import get_status
+
+        # Create a valid store with data
+        from fenris.store import init_store
+        conn = init_store(tmp_path / "test.db")
+        conn.execute(
+            "INSERT INTO samples (ts, device, data_units_written, data_units_read, "
+            "percentage_used, bytes_written, bytes_read, power_on_hours) "
+            "VALUES (?, ?, ?, ?, ?, ?, ?, ?)",
+            ("2026-09-30T10:00:00+00:00", "/dev/nvme0n1", 1000000, 500000,
+             5, 512000000000, 256000000000, 8765),
+        )
+        conn.commit()
+        conn.close()
+
+        # Save non-default preferences
+        config_home = _make_prefs_dir(tmp_path)
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            save_preferences(theme="high_contrast", reduced_motion=True)
+            status_text = get_status(
+                store_path=tmp_path / "test.db",
+                clock_now=datetime(2026, 9, 30, 12, 0, 0, tzinfo=timezone.utc),
+                query_services=False,
+                query_journal=False,
+            )
+            # Status output should NOT contain theme names or motion settings
+            assert "high_contrast" not in status_text.lower()
+            assert "reduced_motion" not in status_text.lower()
+            assert "amber" not in status_text.lower()
+
+    def test_preferences_do_not_alter_store(self, tmp_path):
+        """Saving preferences never writes to the observation store."""
+        import sqlite3
+        config_home = _make_prefs_dir(tmp_path)
+        store_path = tmp_path / "observations.db"
+
+        from fenris.store import init_store
+        conn = init_store(store_path)
+        # Record store state before preference save
+        cursor = conn.execute("SELECT name FROM sqlite_master WHERE type='table'")
+        tables_before = sorted(r[0] for r in cursor.fetchall())
+        conn.close()
+
+        with patch.dict(os.environ, {"XDG_CONFIG_HOME": str(config_home)}):
+            save_preferences(theme="nord", reduced_motion=True)
+
+        # Store schema should be unchanged
+        conn = sqlite3.connect(store_path)
+        cursor = conn.execute("SELECT name FROM sqlite_master WHERE type='table'")
+        tables_after = sorted(r[0] for r in cursor.fetchall())
+        conn.close()
+        assert tables_before == tables_after
+
+
+# Need datetime for CLI isolation test
+from datetime import datetime, timezone

tests/test_projection.py

@@ -22,7 +22,7 @@ from fenris.monitoring_periods import ensure_period_open, close_period
 from fenris.projection import (
     compute_projection, ConfidenceState, BaselineTier, ScenarioRange,
     TBW_TO_BYTES, HORIZON_DAYS, WARMING_MIN_DAYS, STALENESS_HOURS,
-    YOUNG_REGIME_DAYS, DISCLOSURES,
+    YOUNG_REGIME_DAYS, DISCLOSURES, _compute_horizon_rate,
 )
 
 
@@ -899,3 +899,126 @@ class TestIdentityChangeBlankKeys:
         # All 25 days in same segment (equal blanks continue)
         assert result.regime_days is not None
         assert result.regime_days >= 20  # Most of the history
+
+
+# ===========================================================================
+# Issue #76: Evidence-anchored projection rates
+# Scenario windows anchored at latest evidence endpoint T
+# ===========================================================================
+
+
+class TestEvidenceAnchoredHorizons:
+    """Issue #76: Scenario windows anchored at latest published usage-evidence
+    endpoint T with exact trailing 7/28/90×86400-second starts."""
+
+    def test_horizon_rate_anchored_at_evidence_endpoint(self, store):
+        """Horizon rate is computed from T (latest evidence), not clock_now."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-01T00:00:00+00:00")
+        _open_period(store, start="2026-09-01T00:00:00+00:00")
+        bw = 100 * 1024 * 1024
+        # 30 days of data ending Sep 29
+        for i in range(30):
+            d = (datetime(2026, 9, 1) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(store, d, bw=bw)
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        # Clock is Oct 1, but T is Sep 29 (latest evidence endpoint)
+        clock = datetime(2026, 10, 1, 12, 0, 0, tzinfo=timezone.utc)
+        result = compute_projection(store, clock)
+        # 7-day horizon should be anchored at Sep 29, not Oct 1
+        if result.scenario_range and 7 in result.scenario_range.rates:
+            # Rate should be based on Sep 23-29, not Sep 25-Oct 1
+            assert result.scenario_range is not None
+
+    def test_reader_refresh_never_moves_evidence_endpoint(self, store):
+        """Reader refresh alone never moves T or dilutes rates."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-01T00:00:00+00:00")
+        _open_period(store, start="2026-09-01T00:00:00+00:00")
+        bw = 100 * 1024 * 1024
+        for i in range(30):
+            d = (datetime(2026, 9, 1) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(store, d, bw=bw)
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        # Two reads at different clock times
+        clock1 = datetime(2026, 9, 30, 12, 0, 0, tzinfo=timezone.utc)
+        clock2 = datetime(2026, 10, 1, 12, 0, 0, tzinfo=timezone.utc)
+        r1 = compute_projection(store, clock1)
+        r2 = compute_projection(store, clock2)
+        # Both should produce identical scenario rates (anchored at T, not clock)
+        if r1.scenario_range and r2.scenario_range:
+            assert r1.scenario_range.rates == r2.scenario_range.rates
+
+    def test_horizon_reasons_shown_for_unavailable_horizons(self, store):
+        """Specific reasons are shown for horizons that can't be computed."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-20T00:00:00+00:00")
+        _open_period(store, start="2026-09-20T00:00:00+00:00")
+        bw = 100 * 1024 * 1024
+        # Only 10 days of data
+        for i in range(10):
+            d = (datetime(2026, 9, 20) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(store, d, bw=bw)
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        result = compute_projection(store, _clock())
+        # 7-day horizon should be available, 28 and 90 should have reasons
+        if result.scenario_range:
+            assert 7 in result.scenario_range.rates
+            if 28 in result.scenario_range.horizon_reasons:
+                assert "starts before earliest data" in result.scenario_range.horizon_reasons[28]
+            if 90 in result.scenario_range.horizon_reasons:
+                assert "starts before earliest data" in result.scenario_range.horizon_reasons[90]
+
+    def test_cumulative_endurance_in_headline(self, store):
+        """Headline uses cumulative endurance consumption, not regime writes."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-01T00:00:00+00:00")
+        _open_period(store, start="2026-09-01T00:00:00+00:00")
+        bw = 100 * 1024 * 1024
+        for i in range(30):
+            d = (datetime(2026, 9, 1) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(store, d, bw=bw)
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        result = compute_projection(store, _clock())
+        # Headline should be computed with cumulative bytes
+        if result.headline_remaining_seconds is not None:
+            # E_baseline = 10 TB = 10e12 bytes
+            # cumulative_bytes = 30 * 100 * 1024 * 1024
+            # rate = cumulative_bytes / wall_clock
+            # headline = (E_baseline - cumulative_bytes) / rate
+            E_baseline = 10.0 * TBW_TO_BYTES
+            cumulative_bytes = 30 * bw
+            assert result.headline_remaining_seconds >= 0
+
+    def test_zero_boundary_delta_returns_zero(self, store):
+        """Zero monotonic delta proves zero over its represented subspan."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-01T00:00:00+00:00")
+        _open_period(store, start="2026-09-01T00:00:00+00:00")
+        # Days with zero bytes written
+        for i in range(30):
+            d = (datetime(2026, 9, 1) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(store, d, bw=0)
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        result = compute_projection(store, _clock())
+        # Zero rate should result in UNSUPPORTED
+        assert result.confidence_state == ConfidenceState.UNSUPPORTED
+        assert result.headline_remaining_seconds is None
+
+    def test_noon_endpoint_same_as_midnight(self, store):
+        """Noon endpoint produces same rates as midnight endpoint."""
+        _insert_baseline(store, tbw_tb=10.0, verified=True)
+        _insert_segment(store, opened_at="2026-09-01T00:00:00+00:00")
+        _open_period(store, start="2026-09-01T00:00:00+00:00")
+        bw = 100 * 1024 * 1024
+        for i in range(30):
+            d = (datetime(2026, 9, 1) + timedelta(days=i)).strftime("%Y-%m-%d")
+            _insert_day(store, d, bw=bw)
+        _insert_sample(store, "2026-09-30T10:00:00+00:00", pu=5)
+        # Both reads should produce same scenario rates
+        clock1 = datetime(2026, 9, 30, 0, 0, 0, tzinfo=timezone.utc)
+        clock2 = datetime(2026, 9, 30, 12, 0, 0, tzinfo=timezone.utc)
+        r1 = compute_projection(store, clock1)
+        r2 = compute_projection(store, clock2)
+        if r1.scenario_range and r2.scenario_range:
+            assert r1.scenario_range.rates == r2.scenario_range.rates

tests/test_pruning.py

@@ -1,7 +1,9 @@
 """Raw sample pruning tests.
 
 Spec §3.4, ST-5: Raw samples pruned to 14 days; hour observations and
-day aggregates retained indefinitely.
+day aggregates are retained indefinitely.
+
+Issue #74: Boundary anchors required for successor evidence are retained.
 """
 import sqlite3
 import sys
@@ -51,6 +53,9 @@ class TestPruneOldSamples:
     def test_removes_old_samples(self, store_conn):
         now = datetime(2026, 9, 15, 12, 0, 0, tzinfo=timezone.utc)
         # Insert samples at 10, 14, and 15 days ago
+        # All three are before the cutoff (2026-09-01T12:00:00)
+        # The 15-day-old sample is not a boundary anchor because
+        # the next sample (14 days ago) is also before the cutoff
         for days_ago in [10, 14, 15]:
             ts = (now - timedelta(days=days_ago)).isoformat()
             _insert_sample(store_conn, ts)
@@ -82,6 +87,7 @@ class TestPruneOldSamples:
         """Hour observations are retained indefinitely."""
         now = datetime(2026, 9, 15, 12, 0, 0, tzinfo=timezone.utc)
         # Insert an old sample and a recent sample
+        # The old sample is a boundary anchor (needed for derivation)
         _insert_sample(store_conn, (now - timedelta(days=20)).isoformat())
         _insert_sample(store_conn, (now - timedelta(days=1)).isoformat())
 
@@ -95,10 +101,55 @@ class TestPruneOldSamples:
 
         prune_old_samples(store_conn, now, retention_days=14)
 
-        # Sample removed
+        # Old sample is retained as boundary anchor (needed for derivation)
         cursor = store_conn.execute("SELECT COUNT(*) FROM samples")
-        assert cursor.fetchone()[0] == 1
+        assert cursor.fetchone()[0] == 2
 
         # Hour observation retained
         cursor = store_conn.execute("SELECT COUNT(*) FROM hour_observations")
         assert cursor.fetchone()[0] == 1
+
+    def test_boundary_anchor_retained(self, store_conn):
+        """Boundary anchors required for derivation are retained."""
+        now = datetime(2026, 9, 30, 12, 0, 0, tzinfo=timezone.utc)
+        
+        # Old sample before boundary (2026-09-14T23:55:00)
+        # is 15 days and 0.75 hours old (before cutoff at 2026-09-16T12:00:00)
+        _insert_sample(store_conn, "2026-09-14T23:55:00+00:00", 1000000)
+        
+        # Sample after boundary (2026-09-16T12:30:00)
+        # is 13 days and 23.5 hours old (within retention)
+        _insert_sample(store_conn, "2026-09-16T12:30:00+00:00", 2000000)
+        
+        # Run pruning
+        pruned = prune_old_samples(store_conn, now, retention_days=14)
+        
+        # The boundary anchor should be retained
+        cursor = store_conn.execute(
+            "SELECT COUNT(*) FROM samples WHERE ts = '2026-09-14T23:55:00+00:00'"
+        )
+        assert cursor.fetchone()[0] == 1
+
+    def test_old_sample_with_derived_interval_removed(self, store_conn):
+        """Old samples with fully derived intervals are removed."""
+        now = datetime(2026, 9, 30, 12, 0, 0, tzinfo=timezone.utc)
+        
+        # Old sample with derived interval
+        _insert_sample(store_conn, "2026-09-10T10:00:00+00:00", 1000000)
+        _insert_sample(store_conn, "2026-09-10T10:30:00+00:00", 2000000)
+        
+        # Hour observation exists for the interval
+        store_conn.execute(
+            "INSERT INTO hour_observations (hour, active_seconds, bytes_written_delta, bytes_read_delta, sample_count, coverage) "
+            "VALUES ('2026-09-10T10:00:00+00:00', 3600, 1000000, 0, 1, 1.0)",
+        )
+        store_conn.commit()
+        
+        # Run pruning
+        pruned = prune_old_samples(store_conn, now, retention_days=14)
+        
+        # Old sample should be removed (interval is derived)
+        cursor = store_conn.execute(
+            "SELECT COUNT(*) FROM samples WHERE ts = '2026-09-10T10:00:00+00:00'"
+        )
+        assert cursor.fetchone()[0] == 0

tests/test_release.py

@@ -0,0 +1,390 @@
+"""Release flow tests (issue #52).
+
+Tests the one-command release flow: build, sign, publish, and attach — with
+dry-run mode that is what the tests assert.  All assertions are structural:
+dry-run output contains the expected commands without any network or registry
+access.
+
+Requirements:
+  - scripts/release.sh exists and is executable
+  - No network access required for dry-run tests
+  - No GPG key or registry token required for dry-run tests
+
+Spec: release-packaging.md §5, issue #52 acceptance criteria
+"""
+import subprocess
+from pathlib import Path
+
+import pytest
+
+REPO_ROOT = Path(__file__).resolve().parent.parent
+RELEASE_SCRIPT = REPO_ROOT / "scripts" / "release.sh"
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+def _read(path: str | Path) -> str:
+    from tests.conftest import read
+    return read(path)
+
+
+def _get_version() -> str:
+    """Extract version from pyproject.toml."""
+    from tests.conftest import get_version
+    return get_version()
+
+
+def _run_dry_run(*args: str) -> tuple[int, str]:
+    """Run the release script in dry-run mode and return (exit_code, stdout)."""
+    cmd = ["bash", str(RELEASE_SCRIPT), "--dry-run"] + list(args)
+    r = subprocess.run(
+        cmd, capture_output=True, text=True, timeout=30,
+        cwd=REPO_ROOT,
+    )
+    return r.returncode, r.stdout + r.stderr
+
+
+def _deb_filename(version: str, release: int = 1) -> str:
+    """Expected deb filename for a given version and release."""
+    return f"fenris_{version}_amd64.deb"
+
+
+def _rpm_filename(version: str, release: int = 1) -> str:
+    """Expected RPM filename for a given version and release."""
+    return f"fenris-{version}-{release}.x86_64.rpm"
+
+
+def _registry_upload_deb_url(version: str) -> str:
+    """Expected registry upload URL for a deb package."""
+    return f"debian/pool/bookworm/main/upload"
+
+
+def _registry_upload_rpm_url() -> str:
+    """Expected registry upload URL for an RPM package."""
+    return "rpm/fenris/upload"
+
+
+# ---------------------------------------------------------------------------
+# Tests — release script existence and permissions
+# ---------------------------------------------------------------------------
+
+class TestReleaseScriptExists:
+    """Verify the release script is present and executable."""
+
+    def test_script_exists(self):
+        assert RELEASE_SCRIPT.exists(), \
+            "scripts/release.sh must exist"
+
+    def test_script_is_executable(self):
+        assert RELEASE_SCRIPT.stat().st_mode & 0o111, \
+            "scripts/release.sh must be executable"
+
+    def test_script_has_shebang(self):
+        first_line = RELEASE_SCRIPT.read_text().splitlines()[0]
+        assert first_line.startswith("#!/"), \
+            "scripts/release.sh must have a shebang"
+
+
+# ---------------------------------------------------------------------------
+# Tests — dry-run prints all expected commands
+# ---------------------------------------------------------------------------
+
+class TestDryRunCommandPrintout:
+    """Verify dry-run prints every command that would execute."""
+
+    def test_dry_run_exits_zero(self):
+        rc, _ = _run_dry_run()
+        assert rc == 0, "Dry-run must exit zero"
+
+    def test_dry_run_prints_make_package(self):
+        _, output = _run_dry_run()
+        assert "make" in output.lower() and "package" in output.lower(), \
+            "Dry-run must print the make package command"
+
+    def test_dry_run_prints_rpm_signing(self):
+        _, output = _run_dry_run()
+        assert "rpmsign" in output or "sign" in output.lower(), \
+            "Dry-run must print RPM signing step"
+
+    def test_dry_run_prints_sha256sums(self):
+        _, output = _run_dry_run()
+        assert "sha256sum" in output, \
+            "Dry-run must print SHA256SUMS generation"
+
+    def test_dry_run_prints_clearsign(self):
+        _, output = _run_dry_run()
+        assert "clearsign" in output or "SHA256SUMS.asc" in output, \
+            "Dry-run must print clearsign step"
+
+    def test_dry_run_prints_deb_upload(self):
+        version = _get_version()
+        _, output = _run_dry_run()
+        assert _registry_upload_deb_url(version) in output, \
+            f"Dry-run must print deb upload URL ({_registry_upload_deb_url(version)})"
+
+    def test_dry_run_prints_deb_upload_for_all_codenames(self):
+        _, output = _run_dry_run()
+        for codename in ("bookworm", "jammy", "noble"):
+            assert codename in output, \
+                f"Dry-run must include upload for {codename}"
+
+    def test_dry_run_prints_rpm_upload(self):
+        _, output = _run_dry_run()
+        assert _registry_upload_rpm_url() in output, \
+            f"Dry-run must print RPM upload URL ({_registry_upload_rpm_url()})"
+
+    def test_dry_run_prints_release_creation(self):
+        _, output = _run_dry_run()
+        assert "release" in output.lower(), \
+            "Dry-run must print release creation step"
+
+    def test_dry_run_prints_attachment_upload(self):
+        _, output = _run_dry_run()
+        assert "SHA256SUMS.asc" in output, \
+            "Dry-run must print SHA256SUMS.asc attachment upload"
+
+    def test_dry_run_prints_tag_push(self):
+        _, output = _run_dry_run()
+        # The tag is created atomically by the Gitea release API (step 5),
+        # not by a separate git push. Verify the release creation step is present.
+        assert "tag_name" in output or "release" in output.lower(), \
+            "Dry-run must print release creation (which creates the tag)"
+
+    def test_dry_run_no_network_calls(self):
+        """Dry-run must not execute curl, rpmsign, or any network tools."""
+        _, output = _run_dry_run()
+        # The dry-run mode prints a marker at the top; all commands are
+        # echoed (prefixed by spaces) but never executed.  Verify the
+        # marker is present, confirming we're in dry-run mode.
+        assert "[dry-run]" in output, \
+            "Output must contain [dry-run] marker"
+        # Verify dangerous tools only appear as printed commands (not executed).
+        # Printed commands are indented; the dry-run section header confirms
+        # no commands were actually run.
+        assert "Commands below will be executed" in output, \
+            "Dry-run must indicate commands are for display only"
+
+
+# ---------------------------------------------------------------------------
+# Tests — dry-run prints correct package filenames
+# ---------------------------------------------------------------------------
+
+class TestDryRunFilenames:
+    """Verify dry-run uses the correct artifact filenames."""
+
+    def test_deb_filename_in_output(self):
+        version = _get_version()
+        _, output = _run_dry_run()
+        expected = _deb_filename(version)
+        assert expected in output, \
+            f"Dry-run must reference deb filename {expected}"
+
+    def test_rpm_filename_in_output(self):
+        version = _get_version()
+        _, output = _run_dry_run()
+        expected = _rpm_filename(version)
+        assert expected in output, \
+            f"Dry-run must reference RPM filename {expected}"
+
+    def test_checksums_filename_in_output(self):
+        _, output = _run_dry_run()
+        assert "SHA256SUMS" in output, \
+            "Dry-run must reference SHA256SUMS filename"
+
+
+# ---------------------------------------------------------------------------
+# Tests — dry-run does not create artifacts or tags
+# ---------------------------------------------------------------------------
+
+class TestDryRunNoSideEffects:
+    """Verify dry-run creates no filesystem or git side effects."""
+
+    def test_dry_run_no_git_tag_created(self):
+        version = _get_version()
+        tag = f"v{version}"
+        # Ensure tag doesn't exist before
+        r = subprocess.run(
+            ["git", "tag", "-l", tag], capture_output=True, text=True,
+            cwd=REPO_ROOT,
+        )
+        pre_tags = r.stdout.strip()
+
+        _run_dry_run()
+
+        # Verify tag was not created
+        r = subprocess.run(
+            ["git", "tag", "-l", tag], capture_output=True, text=True,
+            cwd=REPO_ROOT,
+        )
+        post_tags = r.stdout.strip()
+        assert pre_tags == post_tags, \
+            f"Dry-run must not create git tag {tag}"
+
+
+# ---------------------------------------------------------------------------
+# Tests — revision bumping (structural: output contains incremented release)
+# ---------------------------------------------------------------------------
+
+class TestRevisionBumping:
+    """Verify the release script handles revision bumping.
+
+    When a version already exists in the registry (HTTP 409), the script
+    bumps the revision and retries.  These tests verify the dry-run output
+    reflects the correct revision logic — without any network access.
+    """
+
+    def test_dry_run_starts_at_revision_one(self):
+        version = _get_version()
+        _, output = _run_dry_run()
+        rpm_expected = _rpm_filename(version, 1)
+        assert rpm_expected in output, \
+            f"Dry-run must start at release 1: expected {rpm_expected} in output"
+
+    def test_revision_bump_changes_rpm_filename(self):
+        """When revision is bumped, the RPM filename changes accordingly."""
+        version = _get_version()
+        rpm_r1 = _rpm_filename(version, 1)
+        rpm_r2 = _rpm_filename(version, 2)
+        # R2 filename must differ from R1
+        assert rpm_r1 != rpm_r2, \
+            "R2 filename must differ from R1"
+        # Both must contain the version
+        assert version in rpm_r1
+        assert version in rpm_r2
+
+    def test_revision_bump_changes_deb_filename(self):
+        """When revision is bumped, the deb filename also changes."""
+        version = _get_version()
+        # Deb filename includes release in nfpm naming
+        deb_r1 = f"fenris_{version}_amd64.deb"
+        deb_r2 = f"fenris_{version}_amd64.deb"
+        # For deb, the filename doesn't change with revision (deb uses epoch)
+        # But the RPM does — this verifies we test RPM revision correctly
+        rpm_r1 = _rpm_filename(version, 1)
+        rpm_r2 = _rpm_filename(version, 2)
+        assert "-1." in rpm_r1, "R1 RPM must contain -1."
+        assert "-2." in rpm_r2, "R2 RPM must contain -2."
+
+
+# ---------------------------------------------------------------------------
+# Tests — bare tag prevention (structural)
+# ---------------------------------------------------------------------------
+
+class TestBareTagPrevention:
+    """Verify the flow prevents bare tags.
+
+    A bare tag (tag without packages, release entry, notes, and checksums)
+    must not result from the flow.  The script checks for existing bare
+    tags before proceeding.  These tests verify the dry-run doesn't create
+    any tags.
+    """
+
+    def test_dry_run_does_not_push_tag(self):
+        _, output = _run_dry_run()
+        # The dry-run marker confirms no commands are executed.
+        # git push appears only as a printed command, never executed.
+        assert "[dry-run]" in output, \
+            "Must be in dry-run mode"
+        # Tag push is printed but the [dry-run] marker confirms nothing ran
+        assert "Commands below will be executed" in output, \
+            "Dry-run must indicate commands are for display only"
+
+
+# ---------------------------------------------------------------------------
+# Tests — CI workflow file
+# ---------------------------------------------------------------------------
+
+class TestCIWorkflow:
+    """Verify the dormant CI workflow is present and correctly structured."""
+
+    def test_workflow_file_exists(self):
+        path = REPO_ROOT / ".gitea" / "workflows" / "release.yml"
+        assert path.exists(), \
+            ".gitea/workflows/release.yml must exist"
+
+    def test_workflow_triggers_on_tags(self):
+        content = _read(".gitea/workflows/release.yml")
+        assert "v*" in content, \
+            "Workflow must trigger on version tags (v*)"
+
+    def test_workflow_has_release_step(self):
+        content = _read(".gitea/workflows/release.yml")
+        assert "release" in content.lower(), \
+            "Workflow must have a release step"
+
+    def test_workflow_mentions_signing(self):
+        content = _read(".gitea/workflows/release.yml")
+        assert "sign" in content.lower(), \
+            "Workflow must include signing step"
+
+    def test_workflow_mentions_upload(self):
+        content = _read(".gitea/workflows/release.yml")
+        assert "upload" in content.lower() or "publish" in content.lower(), \
+            "Workflow must include upload/publish step"
+
+    def test_workflow_validates_notes_before_publication(self):
+        content = _read(".gitea/workflows/release.yml")
+        assert "scripts/extract_changelog.py" in content, \
+            "Workflow must fail before publication if release notes cannot be extracted"
+        assert "--footer packaging/release-footer.md" in content, \
+            "Workflow must assemble the body from the standing release footer"
+
+    def test_workflow_resynchronizes_existing_release_bodies(self):
+        content = _read(".gitea/workflows/release.yml")
+        assert "scripts/release_request.py" in content, \
+            "Workflow must make the create-versus-update decision through the request seam"
+        assert '"${METHOD}"' in content, \
+            "Workflow must execute the helper-selected create-or-update request"
+        assert 'RELEASE_PATH="$(printf' in content and '\n          PATH="$(printf' not in content, \
+            "Workflow must not overwrite the shell PATH while preparing the request URL"
+
+
+def test_readme_points_consumers_to_release_notes():
+    readme = _read("README.md")
+
+    assert "Per-release notes live on the [releases page]" in readme
+    assert "standing install and verification instructions" in readme
+
+
+# ---------------------------------------------------------------------------
+# Tests — Makefile release targets
+# ---------------------------------------------------------------------------
+
+class TestMakefileReleaseTargets:
+    """Verify the Makefile exposes release-related targets."""
+
+    def _makefile_content(self) -> str:
+        return _read("Makefile")
+
+    def test_release_run_target_exists(self):
+        content = self._makefile_content()
+        assert "release-run:" in content, \
+            "Makefile must have a release-run target"
+
+    def test_release_dry_run_target_exists(self):
+        content = self._makefile_content()
+        assert "release-dry-run:" in content, \
+            "Makefile must have a release-dry-run target"
+
+    def test_release_run_calls_script(self):
+        content = self._makefile_content()
+        assert "release.sh" in content, \
+            "release-run target must call scripts/release.sh"
+
+    def test_release_dry_run_uses_dry_run_flag(self):
+        content = self._makefile_content()
+        # Find the release-dry-run target and verify it passes --dry-run
+        in_target = False
+        for line in content.splitlines():
+            if line.startswith("release-dry-run:"):
+                in_target = True
+                continue
+            if in_target and line.strip():
+                if "--dry-run" in line:
+                    break
+                if not line.startswith("\t"):
+                    break
+        else:
+            pytest.fail("release-dry-run target must pass --dry-run to release.sh")

tests/test_repair.py

@@ -0,0 +1,398 @@
+"""Repair and retention tests (issue #74).
+
+Tests the idempotent, safe repair of hour observations and day aggregates
+from surviving raw samples, boundary anchor retention, and legacy summary
+handling at actual precision.
+"""
+import sqlite3
+import sys
+from datetime import datetime, timedelta, timezone
+from pathlib import Path
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
+
+from fenris.store import init_store
+from fenris.repair import (
+    repair_derivation,
+    is_repair_in_progress,
+    get_repair_status,
+)
+from fenris.pruning import prune_old_samples, needs_boundary_anchor
+
+
+# ---------------------------------------------------------------------------
+# Fixtures
+# ---------------------------------------------------------------------------
+
+@pytest.fixture
+def store_conn(tmp_path: Path):
+    """Create a fresh store for each test."""
+    db_path = tmp_path / "test.db"
+    conn = init_store(db_path)
+    yield conn
+    conn.close()
+
+
+def _insert_sample(conn, ts_iso, bytes_written, bytes_read=0, power_on_hours=100,
+                   device="/dev/nvme0", segment_id=None):
+    """Insert a raw sample."""
+    conn.execute(
+        """INSERT INTO samples 
+           (ts, device, bytes_written, bytes_read, power_on_hours, 
+            data_units_written, data_units_read, segment_id)
+           VALUES (?, ?, ?, ?, ?, ?, ?, ?)""",
+        (ts_iso, device, bytes_written, bytes_read, power_on_hours,
+         bytes_written // 512000, bytes_read // 512000, segment_id),
+    )
+    conn.commit()
+
+
+def _insert_hour(conn, hour_iso, bytes_written_delta=0, active_seconds=3600,
+                 idle_seconds=0, powered_off_seconds=0, unknown_seconds=0,
+                 sample_count=1, coverage=1.0):
+    """Insert an hour observation."""
+    conn.execute(
+        """INSERT INTO hour_observations 
+           (hour, active_seconds, idle_seconds, powered_off_seconds, unknown_seconds,
+            bytes_written_delta, bytes_read_delta, sample_count, coverage)
+           VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)""",
+        (hour_iso, active_seconds, idle_seconds, powered_off_seconds, unknown_seconds,
+         bytes_written_delta, 0, sample_count, coverage),
+    )
+    conn.commit()
+
+
+def _insert_day_aggregate(conn, day, bytes_written_delta=0, coverage=1.0):
+    """Insert a day aggregate."""
+    conn.execute(
+        """INSERT INTO day_aggregates 
+           (day, active_seconds, bytes_written_delta, coverage, sample_count)
+           VALUES (?, 3600, ?, ?, 1)""",
+        (day, bytes_written_delta, coverage),
+    )
+    conn.commit()
+
+
+def _open_period(conn, start_iso, end_iso=None, end_cause=None):
+    """Insert a monitoring period."""
+    conn.execute(
+        """INSERT INTO monitoring_periods (started_at, ended_at, end_cause)
+           VALUES (?, ?, ?)""",
+        (start_iso, end_iso, end_cause),
+    )
+    conn.commit()
+
+
+# ---------------------------------------------------------------------------
+# AC1: Normal collection, legacy import and recovery use same evidence rules
+# ---------------------------------------------------------------------------
+
+class TestRepairUsesSameEvidenceRules:
+    """AC1: Repair derives only surviving supported evidence, transactionally
+    and idempotently."""
+
+    def test_repair_idempotent_on_empty_store(self, store_conn):
+        """Repair on empty store succeeds and does nothing."""
+        result = repair_derivation(store_conn)
+        assert result.ok is True
+        assert result.hours_created == 0
+        assert result.days_created == 0
+
+    def test_repair_idempotent_on_fully_derived(self, store_conn):
+        """Repair on store with existing derived data does not duplicate."""
+        now = datetime(2026, 9, 15, 12, 0, 0, tzinfo=timezone.utc)
+        _open_period(store_conn, "2026-09-14T00:00:00+00:00")
+        
+        # Insert samples that span an hour
+        _insert_sample(store_conn, "2026-09-14T10:00:00+00:00", 1000000)
+        _insert_sample(store_conn, "2026-09-14T10:30:00+00:00", 2000000)
+        
+        # Pre-existing hour observation for the 10:00 hour
+        # This hour observation captures the interval [10:00, 10:30]
+        _insert_hour(store_conn, "2026-09-14T10:00:00+00:00", 
+                     bytes_written_delta=1000000)
+        _insert_day_aggregate(store_conn, "2026-09-14", 
+                             bytes_written_delta=1000000)
+        
+        # Run repair
+        result = repair_derivation(store_conn)
+        
+        # Should not create new observations (already exist)
+        assert result.hours_created == 0
+        assert result.days_created == 0
+        
+        # Verify no duplicates
+        cursor = store_conn.execute(
+            "SELECT COUNT(*) FROM hour_observations WHERE hour = '2026-09-14T10:00:00+00:00'"
+        )
+        assert cursor.fetchone()[0] == 1
+
+    def test_repair_preserves_import_markers(self, store_conn):
+        """Repair does not remove legacy import markers."""
+        # Set legacy import marker
+        store_conn.execute(
+            "INSERT INTO store_metadata (key, value) VALUES ('legacy_imported', 'true')"
+        )
+        store_conn.commit()
+        
+        # Run repair
+        result = repair_derivation(store_conn)
+        
+        # Verify marker preserved
+        cursor = store_conn.execute(
+            "SELECT value FROM store_metadata WHERE key = 'legacy_imported'"
+        )
+        assert cursor.fetchone()[0] == "true"
+
+
+# ---------------------------------------------------------------------------
+# AC2: Rerunning or interrupting repair produces no duplicates
+# ---------------------------------------------------------------------------
+
+class TestRepairIdempotency:
+    """AC2: No duplicated intervals or totals from repeated repair."""
+
+    def test_repair_no_duplicate_hours(self, store_conn):
+        """Running repair twice produces no duplicate hour observations."""
+        _open_period(store_conn, "2026-09-14T00:00:00+00:00")
+        _insert_sample(store_conn, "2026-09-14T10:00:00+00:00", 1000000)
+        _insert_sample(store_conn, "2026-09-14T10:30:00+00:00", 2000000)
+        
+        # First repair
+        result1 = repair_derivation(store_conn)
+        assert result1.hours_created == 1
+        
+        # Second repair
+        result2 = repair_derivation(store_conn)
+        assert result2.hours_created == 0  # No new hours
+        
+        # Verify only one hour observation
+        cursor = store_conn.execute("SELECT COUNT(*) FROM hour_observations")
+        assert cursor.fetchone()[0] == 1
+
+    def test_repair_no_duplicate_days(self, store_conn):
+        """Running repair twice produces no duplicate day aggregates."""
+        _open_period(store_conn, "2026-09-14T00:00:00+00:00")
+        _insert_sample(store_conn, "2026-09-14T10:00:00+00:00", 1000000)
+        _insert_sample(store_conn, "2026-09-14T10:30:00+00:00", 2000000)
+        
+        # First repair
+        result1 = repair_derivation(store_conn)
+        assert result1.days_created == 1
+        
+        # Second repair
+        result2 = repair_derivation(store_conn)
+        assert result2.days_created == 0  # No new days
+        
+        # Verify only one day aggregate
+        cursor = store_conn.execute("SELECT COUNT(*) FROM day_aggregates")
+        assert cursor.fetchone()[0] == 1
+
+    def test_interrupted_repair_preserves_evidence(self, store_conn):
+        """If repair fails, prior valid history is preserved."""
+        _open_period(store_conn, "2026-09-14T00:00:00+00:00")
+        
+        # Insert valid existing data
+        _insert_hour(store_conn, "2026-09-14T10:00:00+00:00", 
+                     bytes_written_delta=500000)
+        _insert_day_aggregate(store_conn, "2026-09-14", 
+                             bytes_written_delta=500000)
+        
+        # Run repair (should succeed but not modify existing valid data)
+        result = repair_derivation(store_conn)
+        assert result.ok is True
+        
+        # Verify existing data preserved
+        cursor = store_conn.execute(
+            "SELECT bytes_written_delta FROM hour_observations "
+            "WHERE hour = '2026-09-14T10:00:00+00:00'"
+        )
+        assert cursor.fetchone()[0] == 500000
+
+
+# ---------------------------------------------------------------------------
+# AC3: Boundary anchor retention
+# ---------------------------------------------------------------------------
+
+class TestBoundaryAnchorRetention:
+    """AC3: Prune samples only after durable derivation; retain anchors."""
+
+    def test_needs_boundary_anchor_sample(self, store_conn):
+        """Sample before 14-day boundary is needed for derivation."""
+        now = datetime(2026, 9, 30, 12, 0, 0, tzinfo=timezone.utc)
+        
+        # Sample just before 14-day boundary (2026-09-15T23:55:00)
+        # is 14 days and 0.75 hours old (before cutoff at 2026-09-16T12:00:00)
+        _insert_sample(store_conn, "2026-09-15T23:55:00+00:00", 1000000)
+        
+        # Sample just after boundary (2026-09-16T12:30:00)
+        # is 13 days and 23.5 hours old (within retention)
+        _insert_sample(store_conn, "2026-09-16T12:30:00+00:00", 2000000)
+        
+        # The sample at 2026-09-15 is a boundary anchor because
+        # the interval spans the retention boundary
+        assert needs_boundary_anchor(store_conn, "2026-09-15T23:55:00+00:00", now)
+
+    def test_not_boundary_anchor_if_fully_derived(self, store_conn):
+        """Sample that's fully derived is not a boundary anchor."""
+        now = datetime(2026, 9, 30, 12, 0, 0, tzinfo=timezone.utc)
+        
+        # Insert sample and fully derive its interval
+        _insert_sample(store_conn, "2026-09-14T10:00:00+00:00", 1000000)
+        _insert_sample(store_conn, "2026-09-14T10:30:00+00:00", 2000000)
+        
+        # Hour observation already exists for this interval
+        _insert_hour(store_conn, "2026-09-14T10:00:00+00:00", 
+                     bytes_written_delta=1000000)
+        
+        # Not a boundary anchor
+        assert not needs_boundary_anchor(store_conn, "2026-09-14T10:00:00+00:00", now)
+
+    def test_pruning_retains_boundary_anchors(self, store_conn):
+        """Pruning keeps samples needed as boundary anchors."""
+        now = datetime(2026, 9, 30, 12, 0, 0, tzinfo=timezone.utc)
+        
+        # Old sample before boundary (2026-09-14T23:55:00)
+        # is 15 days and 0.75 hours old (before cutoff at 2026-09-16T12:00:00)
+        _insert_sample(store_conn, "2026-09-14T23:55:00+00:00", 1000000)
+        
+        # Sample after boundary (2026-09-16T12:30:00)
+        # is 13 days and 23.5 hours old (within retention)
+        _insert_sample(store_conn, "2026-09-16T12:30:00+00:00", 2000000)
+        
+        # Recent sample
+        _insert_sample(store_conn, "2026-09-29T12:00:00+00:00", 3000000)
+        
+        # Run pruning
+        pruned = prune_old_samples(store_conn, now, retention_days=14)
+        
+        # The boundary anchor should be retained
+        cursor = store_conn.execute(
+            "SELECT COUNT(*) FROM samples WHERE ts = '2026-09-14T23:55:00+00:00'"
+        )
+        assert cursor.fetchone()[0] == 1
+
+    def test_pruning_removes_old_sample_with_derived_interval(self, store_conn):
+        """Pruning removes old samples when interval is fully derived."""
+        now = datetime(2026, 9, 30, 12, 0, 0, tzinfo=timezone.utc)
+        
+        # Old sample with derived interval
+        _insert_sample(store_conn, "2026-09-10T10:00:00+00:00", 1000000)
+        _insert_sample(store_conn, "2026-09-10T10:30:00+00:00", 2000000)
+        
+        # Hour observation exists for the interval
+        _insert_hour(store_conn, "2026-09-10T10:00:00+00:00", 
+                     bytes_written_delta=1000000)
+        
+        # Run pruning
+        pruned = prune_old_samples(store_conn, now, retention_days=14)
+        
+        # Old sample should be removed (interval is derived)
+        cursor = store_conn.execute(
+            "SELECT COUNT(*) FROM samples WHERE ts = '2026-09-10T10:00:00+00:00'"
+        )
+        assert cursor.fetchone()[0] == 0
+
+
+# ---------------------------------------------------------------------------
+# AC4: Legacy day-only summaries retain actual precision
+# ---------------------------------------------------------------------------
+
+class TestLegacySummaryPrecision:
+    """AC4: Legacy summaries at actual precision, no interpolation."""
+
+    def test_legacy_summary_not_reconstructed(self, store_conn):
+        """Legacy day-only summaries are not interpolated to hour detail."""
+        # Insert a legacy-style day aggregate without hour observations
+        _insert_day_aggregate(store_conn, "2026-08-01", 
+                             bytes_written_delta=5000000)
+        
+        # Run repair
+        result = repair_derivation(store_conn)
+        
+        # Should not create hour observations for legacy day
+        cursor = store_conn.execute(
+            "SELECT COUNT(*) FROM hour_observations WHERE hour LIKE '2026-08-01%'"
+        )
+        assert cursor.fetchone()[0] == 0
+
+    def test_legacy_summary_no_double_counting(self, store_conn):
+        """Legacy summaries and derived intervals don't double-count."""
+        # Insert legacy day aggregate
+        _insert_day_aggregate(store_conn, "2026-08-01", 
+                             bytes_written_delta=5000000)
+        
+        # Run repair
+        result = repair_derivation(store_conn)
+        
+        # Day aggregate should not be modified
+        cursor = store_conn.execute(
+            "SELECT bytes_written_delta FROM day_aggregates WHERE day = '2026-08-01'"
+        )
+        assert cursor.fetchone()[0] == 5000000
+
+
+# ---------------------------------------------------------------------------
+# AC5: Status distinguishes evidence states
+# ---------------------------------------------------------------------------
+
+class TestStatusEvidenceDistingushing:
+    """AC5: Read-only views distinguish evidence states."""
+
+    def test_repair_status_available(self, store_conn):
+        """Repair status is available for read-only views."""
+        status = get_repair_status(store_conn)
+        assert hasattr(status, 'last_repair')
+        assert hasattr(status, 'repair_in_progress')
+        assert hasattr(status, 'hours_derived')
+        assert hasattr(status, 'days_derived')
+
+    def test_repair_in_progress_flag(self, store_conn):
+        """Repair in progress flag is trackable."""
+        assert not is_repair_in_progress(store_conn)
+
+
+# ---------------------------------------------------------------------------
+# AC6: Migration then collection then reader consumption
+# ---------------------------------------------------------------------------
+
+class TestMigrationCollectionReader:
+    """AC6: End-to-end migration, collection, and reader consumption."""
+
+    def test_store_with_raw_evidence_and_summaries(self, store_conn):
+        """Store with raw evidence and old summaries works correctly."""
+        # Set up store with mixed data
+        _open_period(store_conn, "2026-08-01T00:00:00+00:00")
+        
+        # Old day-only summary (legacy)
+        _insert_day_aggregate(store_conn, "2026-08-01", 
+                             bytes_written_delta=5000000)
+        
+        # Recent raw samples
+        _insert_sample(store_conn, "2026-09-14T10:00:00+00:00", 1000000)
+        _insert_sample(store_conn, "2026-09-14T10:30:00+00:00", 2000000)
+        
+        # Run repair
+        result = repair_derivation(store_conn)
+        assert result.ok is True
+        
+        # Verify legacy summary preserved
+        cursor = store_conn.execute(
+            "SELECT bytes_written_delta FROM day_aggregates WHERE day = '2026-08-01'"
+        )
+        assert cursor.fetchone()[0] == 5000000
+        
+        # Verify new hour observation created
+        cursor = store_conn.execute(
+            "SELECT COUNT(*) FROM hour_observations WHERE hour LIKE '2026-09-14%'"
+        )
+        assert cursor.fetchone()[0] == 1
+
+    def test_concurrent_reader_consistency(self, store_conn):
+        """Reader sees consistent snapshot during repair."""
+        # This is more of a documentation test - SQLite WAL mode handles this
+        # We verify the store is in WAL mode
+        cursor = store_conn.execute("PRAGMA journal_mode")
+        assert cursor.fetchone()[0] == "wal"

tests/test_signing.py

@@ -0,0 +1,480 @@
+"""Signing and consumer-repo structural tests (issue #51).
+
+Verifies that the signing infrastructure, consumer setup docs, and key
+publication are correctly wired — without requiring a real GPG key,
+network access, or Docker.
+
+All assertions are structural: config keys exist, URLs match, docs
+are present, and the Makefile exposes the right targets.  A throwaway
+test key exercise is included for rpm signature verification mechanics.
+"""
+import subprocess
+import tempfile
+from pathlib import Path
+
+import pytest
+
+REPO_ROOT = Path(__file__).resolve().parent.parent
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+def _read(path: str | Path) -> str:
+    from tests.conftest import read
+    return read(path)
+
+
+def _gpg_available() -> bool:
+    try:
+        r = subprocess.run(
+            ["gpg", "--version"], capture_output=True, timeout=5,
+        )
+        return r.returncode == 0
+    except (FileNotFoundError, subprocess.TimeoutExpired):
+        return False
+
+
+def _rpmsign_available() -> bool:
+    try:
+        r = subprocess.run(
+            ["rpmsign", "--version"], capture_output=True, timeout=5,
+        )
+        return r.returncode == 0
+    except (FileNotFoundError, subprocess.TimeoutExpired):
+        return False
+
+
+# ---------------------------------------------------------------------------
+# Tests — nfpm.yaml signing configuration
+# ---------------------------------------------------------------------------
+
+class TestNfpmSigningConfig:
+    """Verify that nfpm.yaml is correctly configured for RPM builds.
+
+    Note: RPM signing is done via rpmsign post-build (make sign-rpm),
+    not through nfpm's built-in signing. This keeps the build unsigned
+    and the signing step explicit and key-controlled.
+    """
+
+    def test_rpm_overrides_exist(self):
+        """nfpm.yaml must have overrides.rpm for per-format deltas."""
+        content = _read("packaging/nfpm.yaml")
+        assert "overrides:" in content, "overrides section missing from nfpm.yaml"
+        assert "rpm:" in content, "rpm overrides missing from nfpm.yaml"
+
+    def test_rpm_scripts_configured(self):
+        """RPM must use the dedicated scriptlets, not deb scripts."""
+        content = _read("packaging/nfpm.yaml")
+        assert "packaging/rpm/post.sh" in content, \
+            "RPM postinstall must use rpm/post.sh"
+        assert "packaging/rpm/preun.sh" in content, \
+            "RPM preremove must use rpm/preun.sh"
+        assert "packaging/rpm/postun.sh" in content, \
+            "RPM postremove must use rpm/postun.sh"
+
+    def test_rpm_depends_use_correct_syntax(self):
+        """RPM dependencies must use rpm-style version syntax."""
+        content = _read("packaging/nfpm.yaml")
+        assert "python3 >= 3.10" in content, \
+            "RPM depends must use rpm-style version constraint"
+
+
+# ---------------------------------------------------------------------------
+# Tests — Makefile signing targets
+# ---------------------------------------------------------------------------
+
+class TestMakefileSigningTargets:
+    """Verify that the Makefile exposes signing-related targets."""
+
+    def _makefile_content(self) -> str:
+        return _read("Makefile")
+
+    def test_generate_test_key_target(self):
+        content = self._makefile_content()
+        assert "generate-test-key:" in content, \
+            "Makefile must have a generate-test-key target"
+        assert "packaging-test@bongbetic.com" in content or \
+               "packaging@bongbetic.com" in content, \
+            "generate-test-key must reference the packaging key UID"
+
+    def test_sign_rpm_target(self):
+        content = self._makefile_content()
+        assert "sign-rpm:" in content, \
+            "Makefile must have a sign-rpm target"
+        assert "rpmsign" in content, \
+            "sign-rpm target must use rpmsign"
+
+    def test_checksums_target(self):
+        content = self._makefile_content()
+        assert "checksums:" in content, \
+            "Makefile must have a checksums target"
+        assert "sha256sum" in content, \
+            "checksums target must use sha256sum"
+
+    def test_clearsign_target(self):
+        content = self._makefile_content()
+        assert "clearsign:" in content, \
+            "Makefile must have a clearsign target"
+        assert "--clearsign" in content, \
+            "clearsign target must use gpg --clearsign"
+
+    def test_release_depends_on_signing(self):
+        content = self._makefile_content()
+        for line in content.splitlines():
+            if line.startswith("release:"):
+                deps = line.split(":", 1)[1].strip()
+                assert "sign-rpm" in deps, \
+                    "release target must depend on sign-rpm"
+                assert "clearsign" in deps, \
+                    "release target must depend on clearsign"
+                break
+        else:
+            pytest.fail("release target not found in Makefile")
+
+    def test_packaging_key_uid_defined(self):
+        content = self._makefile_content()
+        assert "PACKAGING_KEY" in content, \
+            "Makefile must define PACKAGING_KEY variable"
+
+    def test_release_mentions_key_ceremony(self):
+        content = self._makefile_content()
+        assert "signing-key-ceremony.md" in content, \
+            "release target must reference the key ceremony doc"
+
+
+# ---------------------------------------------------------------------------
+# Tests — fenris.repo configuration
+# ---------------------------------------------------------------------------
+
+class TestFenrisRepo:
+    """Verify the dnf repo file is correctly configured for Fenris."""
+
+    def _repo_content(self) -> str:
+        return _read("packaging/fenris.repo")
+
+    def test_gpgcheck_enabled(self):
+        content = self._repo_content()
+        assert "gpgcheck=1" in content, \
+            "fenris.repo must set gpgcheck=1 for payload verification"
+
+    def test_repo_gpgcheck_disabled(self):
+        content = self._repo_content()
+        assert "repo_gpgcheck=0" in content, \
+            "fenris.repo must set repo_gpgcheck=0 (metadata check via TLS)"
+
+    def test_gpgkey_points_to_packaging_key(self):
+        content = self._repo_content()
+        assert "gpgkey=" in content, \
+            "fenris.repo must have a gpgkey directive"
+        assert "fenris-packaging.asc" in content, \
+            "gpgkey must point at the packaging key"
+        assert "raw/branch/main" in content, \
+            "gpgkey must use raw URL for the public key"
+
+    def test_baseurl_is_fenris_rpm_group(self):
+        content = self._repo_content()
+        assert "rpm/fenris" in content, \
+            "baseurl must point at the fenris RPM group"
+
+
+# ---------------------------------------------------------------------------
+# Tests — public key publication
+# ---------------------------------------------------------------------------
+
+class TestKeyPublication:
+    """Verify the public key is published in-repo with correct metadata."""
+
+    def test_key_file_exists(self):
+        key_path = REPO_ROOT / "packaging" / "keys" / "fenris-packaging.asc"
+        assert key_path.exists(), \
+            "packaging/keys/fenris-packaging.asc must exist"
+
+    def test_key_file_has_raw_url(self):
+        content = _read("packaging/keys/fenris-packaging.asc")
+        raw_url = (
+            "https://git.bongbetic.com/xavierk/Fenris/raw/branch/main/"
+            "packaging/keys/fenris-packaging.asc"
+        )
+        assert raw_url in content, \
+            "Key file must contain its own raw URL as documentation"
+
+    def test_key_file_documents_algorithm(self):
+        content = _read("packaging/keys/fenris-packaging.asc")
+        assert "RSA 3072" in content or "rsa3072" in content.lower(), \
+            "Key file must document the algorithm as RSA 3072"
+
+    def test_key_file_documents_uid(self):
+        content = _read("packaging/keys/fenris-packaging.asc")
+        assert "Fenris Packaging" in content, \
+            "Key file must document the UID"
+
+    def test_key_file_documents_expiry(self):
+        content = _read("packaging/keys/fenris-packaging.asc")
+        assert "2 year" in content or "2-year" in content or "expiry" in content.lower(), \
+            "Key file must document the expiry policy"
+
+    def test_key_file_references_ceremony_doc(self):
+        content = _read("packaging/keys/fenris-packaging.asc")
+        assert "signing-key-ceremony.md" in content, \
+            "Key file must reference the key ceremony document"
+
+
+# ---------------------------------------------------------------------------
+# Tests — key ceremony documentation
+# ---------------------------------------------------------------------------
+
+class TestKeyCeremonyDoc:
+    """Verify the key ceremony document is complete and accurate."""
+
+    def _doc_content(self) -> str:
+        return _read("docs/install/signing-key-ceremony.md")
+
+    def test_ceremony_doc_exists(self):
+        assert (REPO_ROOT / "docs" / "install" / "signing-key-ceremony.md").exists(), \
+            "docs/install/signing-key-ceremony.md must exist"
+
+    def test_documents_key_specification(self):
+        content = self._doc_content()
+        assert "RSA 3072" in content, "Must document RSA 3072 algorithm"
+        assert "Fenris Packaging" in content, "Must document the UID"
+        assert "packaging@bongbetic.com" in content, "Must document the email"
+
+    def test_documents_import_sign_delete(self):
+        content = self._doc_content()
+        assert "import" in content.lower(), "Must document import step"
+        assert "sign" in content.lower(), "Must document sign step"
+        assert "delete" in content.lower(), "Must document delete step"
+
+    def test_documents_rotation_outline(self):
+        content = self._doc_content()
+        assert "rotation" in content.lower(), \
+            "Must document key rotation procedure"
+
+    def test_documents_dual_key_approach(self):
+        content = self._doc_content()
+        assert "previous" in content.lower() or "old" in content.lower(), \
+            "Must document old key retention during rotation"
+
+    def test_documents_private_key_storage(self):
+        content = self._doc_content()
+        assert "password manager" in content.lower(), \
+            "Must document that private key lives in password manager"
+
+
+# ---------------------------------------------------------------------------
+# Tests — consumer setup documentation
+# ---------------------------------------------------------------------------
+
+class TestConsumerDocs:
+    """Verify consumer setup docs are present and correctly wired."""
+
+    def _readme_content(self) -> str:
+        return _read("README.md")
+
+    def test_apt_signed_by_flow(self):
+        content = self._readme_content()
+        assert "signed-by" in content, \
+            "README must document apt signed-by keyring flow"
+        assert "keyrings" in content, \
+            "README must show the keyrings directory"
+
+    def test_apt_fingerprint_placeholder(self):
+        content = self._readme_content()
+        assert "Fingerprint" in content or "fingerprint" in content, \
+            "README must include fingerprint placeholder for TOFU hardening"
+
+    def test_dnf_repo_flow(self):
+        content = self._readme_content()
+        assert "dnf config-manager --add-repo" in content or \
+               "dnf install" in content, \
+            "README must document dnf install flow"
+        assert "fenris.repo" in content, \
+            "README must reference the Fenris-owned repo file"
+
+    def test_no_gitea_auto_repo(self):
+        """Gitea's auto-generated .repo must never be referenced in docs."""
+        content = self._readme_content()
+        # The Gitea auto-generated repo would have gpgcheck=1 against the
+        # instance key, which is a trap. Our docs should only reference
+        # our own fenris.repo file.
+        assert "auto-generated" not in content.lower() or \
+               "never" in content.lower(), \
+            "README must not recommend Gitea's auto-generated .repo"
+
+    def test_package_signature_verification(self):
+        content = self._readme_content()
+        assert "rpm -K" in content or "rpm --checksig" in content, \
+            "README must document RPM signature verification"
+        assert "gpg --verify" in content, \
+            "README must document GPG verification for SHA256SUMS"
+
+    def test_migration_from_make_install(self):
+        content = self._readme_content()
+        assert "migrate-from-makeinstall" in content.lower() or \
+               "migration" in content.lower(), \
+            "README must reference the migration runbook"
+
+
+# ---------------------------------------------------------------------------
+# Tests — release spec references
+# ---------------------------------------------------------------------------
+
+class TestReleaseSpecReferences:
+    """Verify the release spec references the ceremony doc and nfpm config."""
+
+    def _spec_content(self) -> str:
+        return _read("docs/spec/release-packaging.md")
+
+    def test_spec_references_rpmsign(self):
+        content = self._spec_content()
+        assert "rpmsign" in content.lower() or "sign-rpm" in content, \
+            "Spec must reference rpmsign or make sign-rpm for RPM signing"
+
+    def test_spec_references_ceremony_doc(self):
+        content = self._spec_content()
+        assert "signing-key-ceremony.md" in content, \
+            "Spec must reference the key ceremony document"
+
+
+# ---------------------------------------------------------------------------
+# Tests — RPM signature mechanics (throwaway test key, no network)
+# ---------------------------------------------------------------------------
+
+class TestRpmSignatureMechanics:
+    """Verify RPM signing mechanics using a throwaway test key.
+
+    These tests generate a temporary GPG key, build an RPM (or use an
+    existing one), sign it, and verify the signature — all without
+    network access.  They require gpg and rpmsign to be available.
+    """
+
+    @pytest.mark.skipif(
+        not _gpg_available() or not _rpmsign_available(),
+        reason="gpg or rpmsign not available",
+    )
+    def test_throwaway_key_signs_and_verifies(self):
+        """Generate a throwaway key, sign a test RPM, verify signature."""
+        # Find existing RPM
+        version = None
+        for line in (REPO_ROOT / "pyproject.toml").read_text().splitlines():
+            if line.startswith("version"):
+                version = line.split("=")[1].strip().strip('"')
+                break
+        rpm_path = REPO_ROOT / "dist" / f"fenris-{version}-1.x86_64.rpm"
+        if not rpm_path.exists():
+            pytest.skip("RPM not built — run `make package-rpm` first")
+
+        key_uid = "fenris-test-signing@example.com"
+        try:
+            # Generate throwaway key
+            subprocess.run(
+                ["gpg", "--batch", "--gen-key"],
+                input=f"""%no-protection
+Key-Type: RSA
+Key-Length: 3072
+Name-Real: {key_uid}
+Name-Email: {key_uid}
+Expire-Date: 0
+%commit
+""",
+                text=True, check=True, timeout=30,
+            )
+
+            # Copy RPM to temp dir for signing
+            with tempfile.TemporaryDirectory() as tmpdir:
+                signed_rpm = Path(tmpdir) / rpm_path.name
+                signed_rpm.write_bytes(rpm_path.read_bytes())
+
+                # Sign the RPM
+                subprocess.run(
+                    ["rpmsign", "--addsign",
+                     "--define", f"_gpg_name {key_uid}",
+                     str(signed_rpm)],
+                    check=True, timeout=30,
+                )
+
+                # Verify the signature exists and has correct format
+                # (rpm -Kv returns NOKEY if key isn't imported, but the
+                # signature header is still present and verifiable)
+                r = subprocess.run(
+                    ["rpm", "-Kv", str(signed_rpm)],
+                    capture_output=True, text=True, timeout=10,
+                )
+                output = r.stdout + r.stderr
+                assert "RSA" in output or "rsa" in output.lower(), \
+                    f"RPM must have RSA signature: {output}"
+                assert "SHA256" in output or "sha256" in output.lower(), \
+                    f"RPM must have SHA256 digest: {output}"
+                assert "Header V4" in output or "Header" in output, \
+                    f"RPM must have V4 signature header: {output}"
+                assert "Signature" in output, \
+                    f"RPM must show signature info: {output}"
+
+        finally:
+            # Clean up the test key
+            subprocess.run(
+                ["gpg", "--batch", "--yes", "--delete-secret-keys", key_uid],
+                capture_output=True, timeout=5,
+            )
+            subprocess.run(
+                ["gpg", "--batch", "--yes", "--delete-keys", key_uid],
+                capture_output=True, timeout=5,
+            )
+
+    @pytest.mark.skipif(
+        not _gpg_available(),
+        reason="gpg not available",
+    )
+    def test_clearsign_and_verify(self):
+        """Clearsign a test manifest and verify the signature."""
+        key_uid = "fenris-test-clearsign@example.com"
+        try:
+            # Generate throwaway key
+            subprocess.run(
+                ["gpg", "--batch", "--gen-key"],
+                input=f"""%no-protection
+Key-Type: RSA
+Key-Length: 3072
+Name-Real: {key_uid}
+Name-Email: {key_uid}
+Expire-Date: 0
+%commit
+""",
+                text=True, check=True, timeout=30,
+            )
+
+            with tempfile.TemporaryDirectory() as tmpdir:
+                sums = Path(tmpdir) / "SHA256SUMS"
+                sums.write_text(
+                    "abc123  fenris_0.3.0_amd64.deb\n"
+                    "def456  fenris-0.3.0-1.x86_64.rpm\n"
+                )
+
+                # Clearsign
+                subprocess.run(
+                    ["gpg", "--batch", "--yes", "--clearsign",
+                     "--local-user", key_uid, str(sums)],
+                    check=True, timeout=10,
+                )
+
+                # Verify (clearsigned file — just one argument to --verify)
+                r = subprocess.run(
+                    ["gpg", "--verify", str(sums.with_suffix(".asc"))],
+                    capture_output=True, text=True, timeout=10,
+                )
+                assert r.returncode == 0, \
+                    f"Clearsign verification failed: {r.stderr}"
+                assert "Good signature" in r.stderr, \
+                    f"Expected Good signature: {r.stderr}"
+
+        finally:
+            subprocess.run(
+                ["gpg", "--batch", "--yes", "--delete-secret-keys", key_uid],
+                capture_output=True, timeout=5,
+            )
+            subprocess.run(
+                ["gpg", "--batch", "--yes", "--delete-keys", key_uid],
+                capture_output=True, timeout=5,
+            )

tests/test_status.py

@@ -415,6 +415,131 @@ class TestServiceFacts:
         assert "last collect:" in result
         assert "freshness:" in result
 
+    def test_continuity_reports_boot_enabled_independently_of_runtime(self, tmp_path):
+        """Status names reboot continuity while retaining the timer fact."""
+        from fenris.status import get_status
+
+        db = tmp_path / "observations.db"
+        init_store(db)
+        now = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+
+        with patch("fenris.status.query_service_state", return_value={
+            "boot_enabled": True, "timer_active": False,
+            "last_collect_ok": None, "last_collect_age_s": None,
+            "last_collect_reason": None,
+        }):
+            result = get_status(store_path=db, clock_now=now,
+                                query_services=True, query_journal=False)
+
+        assert "monitoring: active in background · persists across reboots" in result
+        assert "timer: inactive" in result
+
+    def test_continuity_and_deliberate_pause_are_reported_separately(self, tmp_path):
+        """Only a sanctioned user_disabled period renders the paused wording."""
+        from fenris.status import get_status
+
+        db = tmp_path / "observations.db"
+        conn = init_store(db)
+        conn.execute(
+            "INSERT INTO monitoring_periods (started_at, ended_at, end_cause) "
+            "VALUES (?, ?, ?)",
+            ("2026-09-01T09:00:00+00:00", "2026-09-01T10:00:00+00:00", "user_disabled"),
+        )
+        conn.commit()
+        conn.close()
+        now = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+
+        with patch("fenris.status.query_service_state", return_value={
+            "boot_enabled": False, "timer_active": False,
+            "last_collect_ok": None, "last_collect_age_s": None,
+            "last_collect_reason": None,
+        }):
+            result = get_status(store_path=db, clock_now=now,
+                                query_services=True, query_journal=False)
+
+        assert "monitoring: does not start on next boot" in result
+        assert "monitoring: paused — deliberate disable" in result
+        assert "paused time is excluded from your usage habit · resume: fenris monitor resume" in result
+
+    def test_raw_system_state_without_user_disabled_is_not_a_deliberate_pause(self, tmp_path):
+        """A non-sanctioned stop never acquires the deliberate-disable label."""
+        from fenris.status import get_status
+
+        db = tmp_path / "observations.db"
+        conn = init_store(db)
+        conn.execute(
+            "INSERT INTO monitoring_periods (started_at, ended_at, end_cause) "
+            "VALUES (?, ?, ?)",
+            ("2026-09-01T09:00:00+00:00", "2026-09-01T10:00:00+00:00", "migrated"),
+        )
+        conn.commit()
+        conn.close()
+        now = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+
+        with patch("fenris.status.query_service_state", return_value={
+            "boot_enabled": False, "timer_active": False,
+            "last_collect_ok": None, "last_collect_age_s": None,
+            "last_collect_reason": None,
+        }):
+            result = get_status(store_path=db, clock_now=now,
+                                query_services=True, query_journal=False)
+
+        assert "monitoring: paused — deliberate disable" not in result
+
+    def test_resumed_open_period_clears_a_previous_deliberate_pause(self, tmp_path):
+        """A later sanctioned resume takes precedence over an older pause."""
+        from fenris.status import get_status
+
+        db = tmp_path / "observations.db"
+        conn = init_store(db)
+        conn.execute(
+            "INSERT INTO monitoring_periods (started_at, ended_at, end_cause) "
+            "VALUES (?, ?, ?)",
+            ("2026-09-01T09:00:00+00:00", "2026-09-01T10:00:00+00:00", "user_disabled"),
+        )
+        conn.execute(
+            "INSERT INTO monitoring_periods (started_at) VALUES (?)",
+            ("2026-09-01T11:00:00+00:00",),
+        )
+        conn.commit()
+        conn.close()
+        now = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+
+        with patch("fenris.status.query_service_state", return_value={
+            "boot_enabled": True, "timer_active": True,
+            "last_collect_ok": None, "last_collect_age_s": None,
+            "last_collect_reason": None,
+        }):
+            result = get_status(store_path=db, clock_now=now,
+                                query_services=True, query_journal=False)
+
+        assert "monitoring: paused — deliberate disable" not in result
+
+    def test_live_enabled_service_suppresses_a_stale_pause_marker(self, tmp_path):
+        """A raw re-enable cannot leave a contradictory paused presentation."""
+        from fenris.status import get_status
+
+        db = tmp_path / "observations.db"
+        conn = init_store(db)
+        conn.execute(
+            "INSERT INTO monitoring_periods (started_at, ended_at, end_cause) "
+            "VALUES (?, ?, ?)",
+            ("2026-09-01T09:00:00+00:00", "2026-09-01T10:00:00+00:00", "user_disabled"),
+        )
+        conn.commit()
+        conn.close()
+        now = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+
+        with patch("fenris.status.query_service_state", return_value={
+            "boot_enabled": True, "timer_active": True,
+            "last_collect_ok": None, "last_collect_age_s": None,
+            "last_collect_reason": None,
+        }):
+            result = get_status(store_path=db, clock_now=now,
+                                query_services=True, query_journal=False)
+
+        assert "monitoring: paused — deliberate disable" not in result
+
 
 # ---------------------------------------------------------------------------
 # Status output structure (§8.8, LC-9)
@@ -441,6 +566,29 @@ class TestStatusOutput:
         assert isinstance(result, str)
         assert len(result) > 0
 
+    def test_status_excludes_tui_identity_and_auth_notice(self, tmp_path):
+        """CLI status never renders TUI-only identity or launch guidance."""
+        from fenris.status import get_status
+
+        db = tmp_path / "observations.db"
+        init_store(db)
+        now = datetime(2026, 9, 1, 12, 0, 0, tzinfo=timezone.utc)
+
+        with patch("fenris.status.query_service_state", return_value={
+            "boot_enabled": False, "timer_active": False,
+            "last_collect_ok": None, "last_collect_age_s": None,
+            "last_collect_reason": None,
+        }):
+            result = get_status(store_path=db, clock_now=now,
+                                query_services=True, query_journal=False)
+
+        for tui_only in (
+            "Fenris — NVMe endurance monitor",
+            "by Bongbetic",
+            "privileged actions will prompt for authentication (polkit)",
+        ):
+            assert tui_only not in result
+
     def test_status_never_writes(self, tmp_path):
         """Status never writes to the store."""
         from fenris.status import get_status

tests/test_store_config.py

@@ -0,0 +1,56 @@
+"""Store path resolution from config — regression coverage for issue #53.
+
+A fresh install ships a placeholder-commented config whose only required
+key is the device selector. The collector must not crash with
+KeyError 'store_path' when the key is absent.
+"""
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
+
+from fenris.store import DEFAULT_STORE_PATH, get_store_path
+
+
+REPO_ROOT = Path(__file__).resolve().parent.parent
+TEMPLATE = REPO_ROOT / "packaging" / "fenris.conf"
+
+
+def _parse_like_load_config(text: str) -> dict:
+    """Mirror collect.load_config()'s key=value parsing rules."""
+    config = {}
+    for line in text.splitlines():
+        line = line.strip()
+        if not line or line.startswith("#"):
+            continue
+        if "=" in line:
+            key, value = line.split("=", 1)
+            config[key.strip()] = value.strip()
+    return config
+
+
+def test_missing_store_path_falls_back_to_default():
+    """Config with only the device selector resolves to the packaged default."""
+    assert get_store_path({"device": "/dev/nvme0n1"}) == DEFAULT_STORE_PATH
+
+
+def test_explicit_store_path_wins():
+    """An explicit store_path override is honored."""
+    assert get_store_path({"store_path": "/tmp/other.db"}) == Path("/tmp/other.db")
+
+
+def test_packaged_template_yields_collectable_config():
+    """The packaged template, once a device is set, must be collector-ready.
+
+    Reproduces the fresh-install path: parse packaging/fenris.conf the way
+    collect.load_config() does, add the device selector, then resolve the
+    store. Issue #53 made this raise KeyError.
+    """
+    config = _parse_like_load_config(TEMPLATE.read_text())
+    config["device"] = "/dev/nvme0n1"
+    assert get_store_path(config) == DEFAULT_STORE_PATH
+
+
+def test_template_documents_store_path():
+    """The template must mention store_path so admins know it is overridable."""
+    assert "store_path" in TEMPLATE.read_text()

tests/test_store_group_access.py

@@ -0,0 +1,79 @@
+"""Group access to the observation store — regression coverage for issue #54.
+
+Two defects: (1) a non-group user's stat() on the store directory raised
+PermissionError straight through open_store_readonly(), crashing status/TUI
+instead of degrading to the Store fault view; (2) even group members could
+not open the WAL-mode store because root-created sidecars lacked group write
+and the store directory lacked group execute-then-write.
+"""
+import sqlite3
+import sys
+from pathlib import Path
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
+
+from fenris.store import DEFAULT_STORE_PATH, init_store
+from fenris.status import StoreFault, open_store_readonly
+
+
+def test_stat_permission_error_becomes_store_fault(monkeypatch, tmp_path):
+    """stat() denied (non-group user on a 2750 dir) → StoreFault, not crash."""
+    store = tmp_path / "observations.db"
+    store.write_bytes(b"")
+
+    import pathlib
+
+    def denied(self, follow_symlinks=True):
+        raise PermissionError(13, "Permission denied")
+
+    monkeypatch.setattr(pathlib.Path, "exists", denied)
+    with pytest.raises(StoreFault):
+        open_store_readonly(store)
+
+
+def test_connect_failure_becomes_store_fault(tmp_path):
+    """sqlite failures stay wrapped as StoreFault (existing contract)."""
+    garbage = tmp_path / "observations.db"
+    garbage.write_bytes(b"not a database" * 100)
+    with pytest.raises(StoreFault):
+        open_store_readonly(garbage)
+
+
+def test_init_store_leaves_files_group_writable(tmp_path):
+    """Root-created stores must stay readable by WAL readers: db and sidecars
+    need group write after init_store (issue #54)."""
+    store = tmp_path / "observations.db"
+    conn = init_store(store)
+    try:
+        assert (store.stat().st_mode & 0o060) == 0o060, "db not group rw"
+        wal = store.with_name(store.name + "-wal")
+        shm = store.with_name(store.name + "-shm")
+        if wal.exists():
+            assert (wal.stat().st_mode & 0o060) == 0o060, "wal not group rw"
+        if shm.exists():
+            assert (shm.stat().st_mode & 0o060) == 0o060, "shm not group rw"
+    finally:
+        conn.close()
+
+
+def test_readonly_open_works_after_init_store(tmp_path):
+    """The shipped read path opens a store created by init_store."""
+    store = tmp_path / "observations.db"
+    writer = init_store(store)
+    writer.execute("INSERT INTO monitoring_periods (started_at) VALUES ('2026-01-01T00:00:00+00:00')")
+    writer.commit()
+    conn = open_store_readonly(store)
+    assert conn is not None
+    conn.close()
+    writer.close()
+
+
+def test_packaging_ships_group_access():
+    """tmpfiles must create the store dir group-writable; collect unit must
+    keep the umask loose so root-created sidecars stay group-accessible."""
+    repo = Path(__file__).resolve().parent.parent
+    assert "2770" in (repo / "packaging" / "tmpfiles.d" / "fenris.conf").read_text()
+    assert "2750" not in (repo / "packaging" / "tmpfiles.d" / "fenris.conf").read_text()
+    assert "UMask=002" in (repo / "units" / "fenris-collect.service").read_text()

tests/test_store_migration.py

@@ -0,0 +1,195 @@
+"""Observation store migration unit tests (issue #48).
+
+Tests the forward-only migration logic that underpins package upgrade
+semantics: older stores are migrated, current stores pass through, and
+newer stores are refused loudly.
+
+Spec: §3.6, §9.5, §10.2
+"""
+import sqlite3
+import sys
+from pathlib import Path
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
+
+from fenris.store import (
+    SCHEMA_VERSION,
+    init_store,
+    migrate_to_latest,
+)
+from fenris.status import NewerSchema, open_store_readonly
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+def _make_store(path: Path, version: int = 0) -> sqlite3.Connection:
+    """Create a store at *path* with the given user_version."""
+    conn = sqlite3.connect(str(path))
+    conn.execute("PRAGMA journal_mode=WAL")
+    if version == 0:
+        # Fresh DB with no schema — user_version defaults to 0
+        pass
+    else:
+        # Create a minimal schema so the DB is valid, then set version
+        conn.execute("""
+            CREATE TABLE IF NOT EXISTS samples (
+                id INTEGER PRIMARY KEY AUTOINCREMENT,
+                ts TEXT NOT NULL,
+                device TEXT NOT NULL
+            )
+        """)
+        conn.execute(f"PRAGMA user_version={version}")
+    conn.commit()
+    return conn
+
+
+# ---------------------------------------------------------------------------
+# migrate_to_latest
+# ---------------------------------------------------------------------------
+
+class TestMigrateToLatest:
+    """Forward-only migration via migrate_to_latest()."""
+
+    def test_migrates_from_zero(self, tmp_path):
+        """Store at user_version=0 → SCHEMA_VERSION (fresh DB)."""
+        db = tmp_path / "observations.db"
+        _make_store(db, version=0)
+
+        steps = migrate_to_latest(db)
+
+        # SCHEMA_VERSION - 0 = SCHEMA_VERSION migration steps
+        assert steps == SCHEMA_VERSION
+
+        # Verify version was bumped
+        conn = sqlite3.connect(str(db))
+        v = conn.execute("PRAGMA user_version").fetchone()[0]
+        conn.close()
+        assert v == SCHEMA_VERSION
+
+    def test_already_current_returns_zero(self, tmp_path):
+        """Store already at SCHEMA_VERSION → 0 steps applied."""
+        db = tmp_path / "observations.db"
+        conn = _make_store(db, version=SCHEMA_VERSION)
+        conn.close()
+
+        steps = migrate_to_latest(db)
+        assert steps == 0
+
+    def test_refuses_newer_store(self, tmp_path):
+        """Store with user_version > SCHEMA_VERSION → ValueError."""
+        db = tmp_path / "observations.db"
+        _make_store(db, version=SCHEMA_VERSION + 1)
+
+        with pytest.raises(ValueError, match="newer Fenris"):
+            migrate_to_latest(db)
+
+    def test_refuses_much_newer_store(self, tmp_path):
+        """Store several versions ahead → ValueError."""
+        db = tmp_path / "observations.db"
+        _make_store(db, version=SCHEMA_VERSION + 5)
+
+        with pytest.raises(ValueError, match="newer Fenris"):
+            migrate_to_latest(db)
+
+    def test_store_not_corrupted_on_refusal(self, tmp_path):
+        """After refusal, store is unchanged (no silent corruption)."""
+        db = tmp_path / "observations.db"
+        _make_store(db, version=SCHEMA_VERSION + 2)
+
+        with pytest.raises(ValueError):
+            migrate_to_latest(db)
+
+        # Version should be unchanged
+        conn = sqlite3.connect(str(db))
+        v = conn.execute("PRAGMA user_version").fetchone()[0]
+        conn.close()
+        assert v == SCHEMA_VERSION + 2
+
+    def test_idempotent_on_current(self, tmp_path):
+        """Calling migrate_to_latest twice on a current store is safe."""
+        db = tmp_path / "observations.db"
+        _make_store(db, version=SCHEMA_VERSION)
+
+        assert migrate_to_latest(db) == 0
+        assert migrate_to_latest(db) == 0
+
+    def test_migrates_intermediate_version(self, tmp_path):
+        """Store at version SCHEMA_VERSION-1 → 1 step to current."""
+        from fenris.store import SCHEMA_VERSION
+        db = tmp_path / "observations.db"
+        _make_store(db, version=SCHEMA_VERSION - 1)
+        steps = migrate_to_latest(db)
+        assert steps == 1
+
+
+# ---------------------------------------------------------------------------
+# init_store — downgrade refusal
+# ---------------------------------------------------------------------------
+
+class TestInitStoreDowngradeRefusal:
+    """init_store() refuses newer-schema stores."""
+
+    def test_refuses_newer_store(self, tmp_path):
+        """init_store raises ValueError on newer-schema store."""
+        db = tmp_path / "observations.db"
+        _make_store(db, version=SCHEMA_VERSION + 1)
+
+        with pytest.raises(ValueError, match="newer Fenris"):
+            init_store(db)
+
+    def test_store_not_corrupted_on_refusal(self, tmp_path):
+        """After init_store refusal, store is unchanged."""
+        db = tmp_path / "observations.db"
+        _make_store(db, version=SCHEMA_VERSION + 1)
+
+        with pytest.raises(ValueError):
+            init_store(db)
+
+        conn = sqlite3.connect(str(db))
+        v = conn.execute("PRAGMA user_version").fetchone()[0]
+        conn.close()
+        assert v == SCHEMA_VERSION + 1
+
+
+# ---------------------------------------------------------------------------
+# open_store_readonly — downgrade refusal
+# ---------------------------------------------------------------------------
+
+class TestOpenStoreReadonlyDowngradeRefusal:
+    """open_store_readonly() raises NewerSchema on newer-schema stores."""
+
+    def test_raises_newer_schema(self, tmp_path):
+        """Newer store → NewerSchema exception."""
+        db = tmp_path / "observations.db"
+        _make_store(db, version=SCHEMA_VERSION + 1)
+
+        with pytest.raises(NewerSchema) as exc_info:
+            open_store_readonly(db)
+
+        assert exc_info.value.version == SCHEMA_VERSION + 1
+
+    def test_store_not_corrupted(self, tmp_path):
+        """After NewerSchema refusal, store is unchanged."""
+        db = tmp_path / "observations.db"
+        _make_store(db, version=SCHEMA_VERSION + 3)
+
+        with pytest.raises(NewerSchema):
+            open_store_readonly(db)
+
+        conn = sqlite3.connect(str(db))
+        v = conn.execute("PRAGMA user_version").fetchone()[0]
+        conn.close()
+        assert v == SCHEMA_VERSION + 3
+
+    def test_current_store_opens(self, tmp_path):
+        """Store at SCHEMA_VERSION opens without error."""
+        db = tmp_path / "observations.db"
+        _make_store(db, version=SCHEMA_VERSION)
+
+        conn = open_store_readonly(db)
+        assert conn is not None
+        conn.close()

tests/test_themes.py

@@ -0,0 +1,186 @@
+"""Tests for Fenris theme presets (issue #80).
+
+Covers:
+  - Three valid presets: Amber, Nord, High Contrast
+  - Amber is the default with amber graph role
+  - Theme roles for graph rendering (allocated, unallocated, gap, zero, partial)
+  - Status semantic colours/glyphs/text always win over theme
+  - Global action reachability and focus contrast in every preset
+"""
+import pytest
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
+
+from fenris.themes import (
+    THEMES,
+    THEME_NAMES,
+    get_theme,
+    get_graph_colors,
+    STATUS_COLORS,
+)
+
+
+# ---------------------------------------------------------------------------
+# Theme registry tests
+# ---------------------------------------------------------------------------
+
+class TestThemeRegistry:
+    """All three presets are registered with correct names."""
+
+    def test_three_themes_registered(self):
+        """Exactly three themes exist."""
+        assert len(THEMES) == 3
+
+    def test_theme_names(self):
+        """Theme names are amber, nord, high_contrast."""
+        assert THEME_NAMES == {"amber", "nord", "high_contrast"}
+
+    def test_get_theme_valid(self):
+        """get_theme returns a Theme for each valid name."""
+        from textual.theme import Theme
+        for name in THEME_NAMES:
+            theme = get_theme(name)
+            assert isinstance(theme, Theme)
+
+    def test_get_theme_invalid_returns_amber(self):
+        """Unknown theme name returns the amber theme."""
+        theme = get_theme("nonexistent")
+        assert theme.name == "fenris-amber"
+
+
+# ---------------------------------------------------------------------------
+# Amber theme tests (default, amber graph role)
+# ---------------------------------------------------------------------------
+
+class TestAmberTheme:
+    """Amber is the default theme with warm tones."""
+
+    def test_amber_is_dark(self):
+        """Amber is a dark theme."""
+        theme = get_theme("amber")
+        assert theme.dark is True
+
+    def test_amber_primary_is_amber(self):
+        """Primary colour is warm amber."""
+        theme = get_theme("amber")
+        assert "d4a017" in theme.primary.lower() or "amber" in theme.primary.lower()
+
+    def test_amber_has_graph_colors(self):
+        """Amber defines all required graph role variables."""
+        colors = get_graph_colors("amber")
+        assert "allocated" in colors
+        assert "unallocated" in colors
+        assert "gap" in colors
+        assert "zero" in colors
+        assert "partial" in colors
+
+
+# ---------------------------------------------------------------------------
+# Nord theme tests
+# ---------------------------------------------------------------------------
+
+class TestNordTheme:
+    """Nord uses the polar night palette."""
+
+    def test_nord_is_dark(self):
+        """Nord is a dark theme."""
+        theme = get_theme("nord")
+        assert theme.dark is True
+
+    def test_nord_has_graph_colors(self):
+        """Nord defines all required graph role variables."""
+        colors = get_graph_colors("nord")
+        assert "allocated" in colors
+        assert "unallocated" in colors
+        assert "gap" in colors
+        assert "zero" in colors
+        assert "partial" in colors
+
+
+# ---------------------------------------------------------------------------
+# High Contrast theme tests
+# ---------------------------------------------------------------------------
+
+class TestHighContrastTheme:
+    """High Contrast for maximum readability."""
+
+    def test_high_contrast_is_dark(self):
+        """High contrast is a dark theme."""
+        theme = get_theme("high_contrast")
+        assert theme.dark is True
+
+    def test_high_contrast_foreground_is_white(self):
+        """Foreground is pure white for maximum contrast."""
+        theme = get_theme("high_contract") if False else get_theme("high_contrast")
+        assert theme.foreground is not None
+
+    def test_high_contrast_has_graph_colors(self):
+        """High contrast defines all required graph role variables."""
+        colors = get_graph_colors("high_contrast")
+        assert "allocated" in colors
+        assert "unallocated" in colors
+        assert "gap" in colors
+        assert "zero" in colors
+        assert "partial" in colors
+
+
+# ---------------------------------------------------------------------------
+# Status semantic colours override theme (AC80-1)
+# ---------------------------------------------------------------------------
+
+class TestStatusSemanticOverride:
+    """Status semantic colours/glyphs/text always win over theme styling."""
+
+    def test_status_colors_defined(self):
+        """STATUS_COLORS maps each StatusState to a fixed colour."""
+        from fenris.status_composition import StatusState
+        for state in StatusState:
+            assert state.value in STATUS_COLORS
+
+    def test_status_colors_are_not_theme_dependent(self):
+        """Status colours are the same regardless of theme."""
+        from fenris.status_composition import StatusState
+        # These are the canonical status colours — they must not change with theme
+        assert STATUS_COLORS[StatusState.MONITORING.value] == "green"
+        assert STATUS_COLORS[StatusState.ERROR.value] == "red"
+        assert STATUS_COLORS[StatusState.PAUSED.value] == "yellow"
+        assert STATUS_COLORS[StatusState.INTERRUPTED.value] == "red"
+        assert STATUS_COLORS[StatusState.STALE.value] == "red"
+        assert STATUS_COLORS[StatusState.WAITING.value] == "yellow"
+        assert STATUS_COLORS[StatusState.UNKNOWN.value] == "dim"
+        assert STATUS_COLORS[StatusState.COLLECTING.value] == "green"
+
+
+# ---------------------------------------------------------------------------
+# Graph colour role tests (AC80-5)
+# ---------------------------------------------------------------------------
+
+class TestGraphColorRoles:
+    """Theme roles for graph rendering expose distinct colours per preset."""
+
+    def test_all_themes_define_same_roles(self):
+        """Every theme has the same set of graph colour roles."""
+        roles = None
+        for name in THEME_NAMES:
+            colors = get_graph_colors(name)
+            if roles is None:
+                roles = set(colors.keys())
+            else:
+                assert set(colors.keys()) == roles
+
+    def test_graph_roles_are_distinct_across_presets(self):
+        """Different themes produce different graph colour values."""
+        amber = get_graph_colors("amber")
+        nord = get_graph_colors("nord")
+        hc = get_graph_colors("high_contrast")
+        # At least one role should differ between themes
+        assert amber["allocated"] != nord["allocated"] or amber["allocated"] != hc["allocated"]
+
+    def test_zero_role_is_dim(self):
+        """Zero-usage bars use a dim/subtle colour in all themes."""
+        for name in THEME_NAMES:
+            colors = get_graph_colors(name)
+            # Zero should be distinct from allocated
+            assert colors["zero"] != colors["allocated"]

units/fenris-collect.service

@@ -0,0 +1,10 @@
+[Unit]
+Description=Fenris NVMe collection service
+Documentation=https://git.bongbetic.com/xavierk/Fenris
+After=local-fs.target
+
+[Service]
+Type=oneshot
+ExecStart=/usr/libexec/fenris/fenris-collect
+TimeoutStartSec=90
+UMask=002

units/fenris-collect.timer

@@ -0,0 +1,12 @@
+[Unit]
+Description=Fenris collection timer
+Documentation=https://git.bongbetic.com/xavierk/Fenris
+
+[Timer]
+OnBootSec=2min
+OnUnitInactiveSec=5min
+AccuracySec=30s
+Persistent=no
+
+[Install]
+WantedBy=timers.target