feat(#73): publish trustworthy first usage history

- Schema migration 1-2: add segment_id to samples, unattributed bytes to day_aggregates

- Collector now derives hour observations and day aggregates from sample pairs

- Cross-hour deltas tracked as unattributed (no proportional allocation)

- Display states: 0 samples -> awaiting first, 1 sample -> awaiting another

- Monitoring period ensured open on each collection run

- Derivation failures preserve prior history

Closes #73

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
+            critical_warning, segment_id
-        ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+        ) 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 and get segment info
-            write_sample(sample, identity, conn, clock)
+            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/status.py

@@ -331,7 +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 = []
 
@@ -361,6 +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)
@@ -598,6 +609,17 @@ 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:
@@ -637,7 +659,7 @@ 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()

src/fenris/store.py

@@ -12,7 +12,7 @@ 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
@@ -106,7 +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
         )
     """)
     
@@ -141,7 +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
         )
     """)
     
@@ -207,11 +210,25 @@ def _apply_migrations(conn: sqlite3.Connection, current_version: int):
     
     Spec: §3.6, §10.2
     """
-    # Migration 1→2: example placeholder
+    # Migration 1→2: add segment_id provenance to samples,
-    # if current_version < 2:
+    # unattributed byte tracking to day_aggregates (issue #73)
-    #     conn.execute("ALTER TABLE ...")
+    if current_version < 2:
-    #     current_version = 2
+        # Defensive: only ALTER if table exists (handles minimal v1 stores)
-    pass
+        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:
@@ -239,6 +256,14 @@ def migrate_to_latest(store_path: Path) -> int:
         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}")

src/fenris/tui.py

@@ -447,7 +447,24 @@ class FenrisTuiApp(App):
 
     def _render_all_regions(self, conn: sqlite3.Connection) -> None:
         """Render all four regions from live store data."""
+        # --- Sample count for single-sample state (issue #73 AC3) ---
+        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 Exception:
+            sample_count = 0
+            day_count = 0
+
         # --- Headline band (§7.2) ---
+        if sample_count <= 1 and day_count == 0:
+            # Single sample: awaiting another sample
+            self._render_headline(
+                "[bold]Awaiting another sample[/bold]\n\n"
+                "Collecting usage data — the first projection requires at least two samples."
+            )
+        else:
             try:
                 proj = compute_projection(conn, self._clock_now)
                 headline = self._format_headline(proj)

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_store_migration.py

@@ -118,12 +118,12 @@ class TestMigrateToLatest:
         assert migrate_to_latest(db) == 0
 
     def test_migrates_intermediate_version(self, tmp_path):
-        """Store at version 1 with SCHEMA_VERSION=1 → 0 steps (current)."""
+        """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=1)
+        _make_store(db, version=SCHEMA_VERSION - 1)
-        # SCHEMA_VERSION is 1, so version 1 is current
         steps = migrate_to_latest(db)
-        assert steps == 0
+        assert steps == 1
 
 
 # ---------------------------------------------------------------------------