docs: offset investigation journal; strip NUL corruption from CLAUDE.md

Adds docs/offset_investigation.md — a dated journal of the "offset" hardware
fault, in the style of the codec status docs: findings with provenance, dead
ends kept with the reason they died, and per-unit case files.

Contents:
  - base rate 5-6 of 45 units (11-13%) over the DL2 archive, 2018-2026,
    confirming rather than overturning the earlier 2-of-21 estimate
  - the detector, with the rationale for each term and its known blind spot
    (event traces carry real motion, so only trace-dominating offsets show)
  - the bimodality result: relaxing the amplitude floor 11x adds no new units
  - Instantel's own procedure and thresholds from their FAQs 13-0-21 / 12-0-10:
    A/D-mode ">5 counts", the autozero key sequence, and the 2027-2069 X1/X8
    acceptance window that explains the ~10% field success rate of a re-zero
  - four ruled-out hypotheses, each with the evidence that killed it:
    condensation, clipping, the sensor check as a predictor (102 offset events,
    zero failures — a grossly offset unit passes its own self-check), and the
    calibration-timing correlation (confounded, one unit per time bucket)
  - open questions, chiefly whether SUB 0x0E carries the autozero numbers

Cross-referenced from CLAUDE.md and Appendix E of the protocol reference
(whose CRLF line endings are preserved).

Separately: CLAUDE.md had 793 NUL bytes appended after its last line. They
predate this work (present at least as far back as e42956a / v0.21.0) and made
grep treat the file as binary, silently skipping it. Stripped.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HgTe8CamXAHcAmaQ6QNcog

CHANGELOG.md

@@ -6,6 +6,43 @@ All notable changes to seismo-relay are documented here.
 
 ## [Unreleased]
 
+### Fixed
+- **Sub-minute histograms with a partial final block decoded to nothing**
+  (`histogram_codec.detect_multi_interval_stride`). The stride search confirmed
+  itself on a third block header whenever the body was long enough to hold one —
+  but a body can exceed two strides and still contain only two real blocks, because
+  a *partial* final block leaves trailing padding. BE18193 `T193L0XM.CI0H` (51
+  intervals at 2 s = one full 30-interval block plus a 21-interval remainder, in a
+  2787-byte body) therefore had its correct stride of 612 discarded and produced an
+  empty decode. A missing third header now means end-of-stream rather than
+  disqualification; the block-counter check, which is what actually prevents the
+  false positives that once mis-dispatched 9,082 files, is unchanged.
+
+  Found by decoding the full DL2 archive against its preserved Blastware ASCII
+  exports. Across **63,535 unique** histogram binaries the fix recovers **4 files** —
+  `K440HJCN.3C0H` and `K557IF1U.8K0H` (stride 252), `T191HVNP.0S0H` (92) and
+  `T193L0XM.CI0H` (612) — with **zero** files regressed. Verification over all
+  14,340 archive pairs goes 14,337 → 14,338 exact, the only remainder being two
+  series-4 IDF files that belong to a different codec.
+
+  (The DL2 export keeps a byte-identical `Sent/` mirror of its root, so a naive
+  walk double-counts every binary — 127,035 paths are 63,535 distinct files. The
+  ASCII exports are *not* mirrored, so the 14,340 pair count is already distinct.)
+
+  ⚠ Prod stores hold `.h5` files generated before this fix. Those 4 events stay
+  empty until `backfill_sidecars.py` is re-run — not worth a two-hour prod backfill
+  on its own; fold it into the next one.
+
+- **Histogram/waveform twin matching is now interval-based** (`find_twins`). A real
+  trigger is recorded twice — as a triggered waveform (stamped at the trigger instant)
+  and inside the scheduled histogram whose interval contains it (stamped at the 7am/7pm
+  interval start) — so the two twins can be **hours apart**. The old ±5-minute window
+  silently missed them, which broke review propagation (flagging one twin didn't flag its
+  twin). Twins are now matched by same serial + identical `peak_vector_sum` + opposite
+  record type + the waveform falling within the histogram's interval (bounded by the next
+  same-serial histogram). `window_seconds` is retained but ignored. Fixes terra-view #102
+  sub-task 2.
+
 ---
 
 ## v0.26.0 — 2026-08-27

CLAUDE.md

@@ -27,8 +27,11 @@ Read this first when picking the project back up.
 - **After any codec change, regenerate the store** — `backfill_sidecars.py
   --force` then `backfill_event_shape.py`, DB backup first. Stored `.h5` files
   do not update themselves.
-- **Parked:** the "offset" hardware-fault investigation (Appendix E of the
-  protocol reference) pending the multi-year BW archive.
+- **The "offset" hardware fault has its own journal** --
+  `docs/offset_investigation.md`. Base rate settled at 5-6 of 45 units
+  (11-13%) across the DL2 archive; Instantel's own autozero procedure and
+  its 2027-2069 go/no-go window are recorded there. Best open lead is
+  `SUB 0x0E` (unimplemented), which may carry those numbers.
 
 
 When new information about the protocol is discovered, please update the instantel_protocol_reference.md with the findings in addition to this document
@@ -1803,4 +1806,4 @@ body) because writing a dial string may require DLE escaping for embedded contro
 
 To parse BW TX captures: use `bridges/captures/` scripts or adapt the `find_write_frames()` pattern
 in `/tmp/analyze_write_payload.py` — it correctly handles `0x10 0x03` DLE-escaped ETX bytes
-inside write frame data (the naive parser terminates early at the escaped `0x03`).
+inside write frame data (the naive parser terminates early at the escaped `0x03`).

docs/instantel_protocol_reference.md

@@ -3524,6 +3524,11 @@ of them was, for a while.
 
 ### E.1 The "offset" fault
 
+> **The full investigation now lives in `docs/offset_investigation.md`** --
+> base rate, detector definition, per-unit case files, ruled-out hypotheses,
+> and Instantel's own autozero procedure with its 2027-2069 acceptance
+> window.  This appendix is kept as the protocol-side summary.
+
 **Symptom.**  One geophone channel's baseline steps away from zero and
 stays there.  The trace still carries the real AC signal, but it rides
 on a DC pedestal of a few tenths of an in/s.  Operators call this an

docs/offset_investigation.md

@@ -0,0 +1,363 @@
+# The "offset" fault — investigation journal
+
+A running record of the **offset** hardware fault on Instantel Series III
+seismographs: a geophone channel whose trace sits displaced from zero rather
+than centred on it.
+
+This is a *journal*, not a spec. Findings are dated, dead ends are kept with
+the reason they died, and every number says where it came from. When something
+here is superseded, strike it and say why rather than deleting it — the point
+is that a future session can tell what was actually established from what was
+merely believed at the time.
+
+Companion material:
+- `scratch/offset_scan.py` — the detector
+- `scratch/verify_against_ascii.py` — decoder verification harness
+- `docs/instantel_protocol_reference.md` — wire protocol, incl. the
+  unimplemented `SUB 0x0E` this investigation now wants
+
+---
+
+## TL;DR (current state, 2026-08-28)
+
+- **It is real device data, not a decode bug.** Settled early and confirmed
+  against Blastware's own ASCII exports.
+- **Base rate: 5–6 of 45 units (11–13%)** across the full DL2 archive,
+  2018–2026. This *confirms* the earlier 2-of-21 (9.5%) estimate from the much
+  smaller Terra-View DB — survivorship bias from deleted events had **not**
+  concealed a wave of cases.
+- **The fault is bimodal, not a drift continuum.** A unit is either clean or
+  grossly off. Loosening the amplitude threshold 11× adds no new units.
+- **The unit's own sensor check cannot see it.** 102 offset events, zero
+  sensor-check failures. Do not try to use it as a screen.
+- **Cause is still unsettled.** Instantel's autozero fixes the minority of
+  cases; the rest are hardware. We cannot yet tell which is which remotely.
+- **Best open lead:** `SUB 0x0E` (channel sensor data, 8 channels × 10 bytes,
+  unimplemented) may carry the very numbers Instantel says to check against
+  **2027–2069**. Untested.
+
+---
+
+## 1. What the fault looks like
+
+A healthy geophone trace is centred on zero. An offset channel is parked away
+from zero, so the channel **mean approaches its own peak**. In Blastware the
+signature is "parallel lines above or below the zero line" (Instantel's own
+wording).
+
+Consequences observed in the field:
+- The unit can **self-trigger on its own offset** when the displacement exceeds
+  the geo trigger level, producing streams of junk events with no ground
+  motion. Instantel has a separate FAQ for this symptom (13-0-22, *"Unit
+  triggers continuously without activity"*).
+- Recorded PPV for that channel is meaningless while the fault persists.
+
+---
+
+## 2. The detector
+
+Implemented in `scratch/offset_scan.py`. Operates on raw BW binaries only — no
+DB, no sidecars.
+
+```
+for each series-3 waveform binary:
+    decode -> per-channel ADC counts
+    dominant axis = channel with the largest |peak|
+    flag when   |mean| / peak  >  0.70
+          and   |mean|         >= 0.90 x the unit's geo trigger level
+episodes = per-serial runs of flagged events, split on a >12 h gap
+```
+
+Why each term:
+
+| term | purpose |
+|---|---|
+| `\|mean\|/peak > 0.7` | the discriminator. A DC-parked trace has mean ≈ peak. |
+| `\|mean\| >= 0.9 × trigger` | amplitude floor — suppresses quiet traces where mean and peak are both tiny and the ratio is meaningless. |
+| dominant axis only | the fault is per-channel; scoring all three dilutes it. |
+| 12 h episode gap | separates deployments/visits rather than counting events. |
+
+Trigger level comes from a paired `_ASCII.TXT` when one exists, else the
+per-serial median learned from that unit's ASCII files, else 0.2 in/s.
+
+**Known limitation.** Event traces contain real ground motion, so this can only
+see offsets large enough to *dominate* the trace. A mild offset on a real blast
+is invisible. Instantel's A/D-mode check (§5) is the only thing that sees the
+mild end. Our base rate is therefore a **gross-offset** rate.
+
+---
+
+## 3. Archive results (2026-08-28)
+
+Source: DL2 event export, 6,577 **unique** series-3 waveforms, 45 units.
+See [`dl2-archive`](#8-data-and-tooling) for the `Sent/` mirror trap.
+
+**283 suspect events, 15 episodes, 6 of 45 units (13.3%).**
+Excluding BE11007 (§4, likely not an offset at all): **5 of 45 = 11.1%**.
+
+### Threshold sensitivity — the bimodality result
+
+Re-scoring the same corpus at a range of amplitude floors, with two
+ratio cut-offs (1 A/D count = 0.005 in/s, see §5):
+
+| \|offset\| floor | m/p > 0.7 | m/p > 0.9 |
+|---|---|---|
+| 5 cts (0.025 in/s) — *Instantel's own* | 333 ev / 6 units | 279 ev / **5 units** |
+| 10 cts (0.050) | 294 / 6 | 274 / 5 |
+| 20 cts (0.100) | 250 / 5 | 244 / 4 |
+| 40 cts (0.200) | 209 / 5 | 203 / 4 |
+| 80 cts (0.400) | 152 / 4 | 148 / 2 |
+| 160 cts (0.800) | 144 / 2 | 141 / 1 |
+
+Relaxing the floor by 11× (0.27 → 0.025 in/s) adds ~14% more events and **no
+new units**. There is no population of mild offsets hiding below our threshold
+*in event data*. Either a unit is clean or it is grossly off.
+
+---
+
+## 4. Per-unit case files
+
+Ordered by severity. `m/p` medians are on the offending channel.
+
+### BE13117 — one violent day, never again
+`145 / 454 events (32%)`, **1 episode**, 2023-05-04, 6.8 h.
+Offset climbed **0.393 → 1.875 in/s within the episode**. `m/p` median
+**0.996** — the trace is almost pure DC. No recurrence in the rest of its 454
+events. No ASCII files in the archive, so no calibration history.
+
+### BE18438 — recurring, months apart
+`87 / 293 (30%)`, **2 episodes**: 2025-11-15 (1.2 h, n=12, 0.279 → 0.369) and
+2026-02-25 (**28.8 h**, n=75, 0.183 → 0.366). `m/p` median 0.967.
+Clean across all 196 events preceding its 2025-08-12 calibration.
+
+### BE9558 — six years apart
+`38 / 196 (19%)`, **4 episodes**: 2020-02-11 (6.3 h, n=33, but only
+0.068 → 0.086 — very mild), then 2026-04-14, 2026-04-29, 2026-05-04
+(0.28–0.45). `m/p` median 0.919. Calibrated 2026-06-26; 0/7 events flagged
+after, but n=7 is far too small to call it fixed.
+
+### BE12599 — the live case ⚠
+`6 / 77 (8%)`, **6 single-event episodes, one per day at exactly 05:00**,
+2026-08-17 → 2026-08-23. Offset rose 0.383 → 0.565 then fell back to 0.345.
+`m/p` ≈ 0.965, geo trigger 0.3 in/s — **the offset exceeds the trigger level,
+so the unit is triggering on its own fault**. Last calibrated 2025-08-12.
+
+This is the most recent and the most useful: a currently-faulting unit is the
+natural experiment for the re-zero-vs-repair question (§7).
+
+### BE11529 — marginal
+`4 / 99 (4%)`, 1 episode 2025-07-08, 0.4 h, offsets only 0.051 → 0.058 in/s.
+`m/p` median 0.959, so DC-dominated, but the magnitude is near the noise of
+this method. Treat as unconfirmed.
+
+### BE11007 — probably NOT an offset
+`3 / 70 (4%)`, 1 episode 2022-01-17, offsets 5.500 → 6.904 in/s — by far the
+largest. But `m/p` is only **0.719–0.738** against ≥0.9 for every other unit,
+and the peaks are 7.6–9.4 in/s on a 10 in/s range. That reads as a **large
+real blast with asymmetric ground motion**, not a parked trace. Excluded from
+the headline base rate.
+
+---
+
+## 5. Instantel's own procedure and thresholds
+
+From two Instantel technical-support FAQs supplied 2026-08-28
+(answers **13-0-21** *"How to determine offsets"* and **12-0-10** *"Removing
+offsets on an Instantel Series III monitor"*; created 2008/2007, last updated
+2009-03-06).
+
+### Identifying (13-0-21)
+
+1. Create or use an event with the **manual minimum trigger** set for the
+   connected geophone and microphone — i.e. an event that recorded no real data.
+2. Save it and open in Blastware.
+3. An offset shows as **parallel lines above or below the zero line**.
+4. Put the unit in **A/D mode** — on Series III, press and hold `OPTION`, then
+   press `START MONITOR`.
+5. **Display counts higher than 5**, with no vibration or overpressure present,
+   indicate an offset.
+
+### Removing — the autozero (12-0-10)
+
+1. Be in a **quiet area with low vibration**.
+2. Power on the Blastmate III / Minimate Plus.
+3. Connect the geophone and microphone — **LINEAR mic only**.
+   ⚠ *Do not connect an "A" weight microphone, regardless of what the monitor
+   displays.*
+4. Press `Test`.
+5. Wait for the **Sensor Check** results to appear.
+6. Press `OPTION` and `START MONITOR` **simultaneously**.
+7. `Performing Autozero` appears; press `Enter`.
+8. Confirm the sensors are properly connected; press `Enter`.
+9. Wait for the autozero to complete.
+10. Press `Enter` twice → Main Menu, *Ready To Monitor*, offset corrected.
+
+### The go/no-go number — 2027 to 2069
+
+> When you perform an Autozero on any Series III unit, the lists of numbers in
+> the **X1 and X8 gains should all be between 2027 and 2069**. If not, repeat
+> the Autozero. **If the numbers are extremely out of the specified range, then
+> the unit should be sent in for repair.**
+>
+> If this process does not remove the offset problem, return the unit **and
+> sensors** to Instantel for repair.
+
+This is the documented explanation for the field experience (Brian's dad,
+2026-08-28) that **a re-zero works maybe 10% of the time** — the autozero only
+recovers units whose zero reference is still near-correct.
+
+### Scale derivation (inference, well-supported — not proven)
+
+2048 is 12-bit midscale. Our codec's geo full scale is 32000 internal counts =
+10 in/s, with 1 decoder unit = 16 counts = exactly 0.005 in/s
+(`geo-full-scale-is-32000-counts`). ±2000 A/D counts about 2048 therefore maps
+to ±10 in/s at **0.005 in/s per A/D count**. That makes:
+
+- Instantel's ">5 counts" threshold ≈ **0.025 in/s**
+- the 2027–2069 window = **±21 counts = ±0.105 in/s** of tolerated zero error
+
+Consistent and mutually corroborating, but we have not confirmed the A/D-count
+scale directly from a device reading.
+
+---
+
+## 6. Ruled out — keep these dead
+
+### Condensation / humidity — DEAD (2026-08-25)
+Proposed, then killed by its own controls: BE18438 stayed flat across a 10-hour
+overnight gap, and only 2 of 21 units showed the fault while 19 sat in the same
+weather. The apparent "diurnal cycle" was an artifact of binning by hour-of-day
+across two days. See `waveform-dc-offset-is-real-device-data`.
+
+### Clipping as a false-positive source — RULED OUT (2026-08-28)
+A rail-hitting trace would fake an offset (mean → peak). It isn't happening:
+median suspect peak is only **10% of full scale**, p90 is 18.6%. Only BE11007's
+3 events exceed 50% FS, and none reach 98%.
+
+### The sensor check as a predictor — DOES NOT WORK (2026-08-28)
+Tested on 102 offset events across 4 units:
+
+| unit | state | n | failed | median ratio | median freq |
+|---|---|---|---|---|---|
+| BE11529 | offset | 4 | **0** | 3.90 | 7.6 |
+| BE11529 | clean | 14 | 0 | 3.80 | 7.5 |
+| BE12599 | offset | 6 | **0** | 4.00 | 7.4 |
+| BE12599 | clean | 13 | 0 | 4.00 | 7.6 |
+| BE18438 | offset | 87 | **0** | 3.70 | 7.6 |
+| BE18438 | clean | 25 | 0 | 3.80 | 7.5 |
+| BE9558 | offset | 5 | **0** | 3.90 | 7.8 |
+| BE9558 | clean | 44 | 0 | 3.80 | 7.5 |
+
+Zero failures on either side and indistinguishable ratios/frequencies. The
+swing test measures geophone frequency response and damping — it never examines
+DC zero. **A grossly offset unit passes its own self-check.** This is why the
+fault goes unnoticed until somebody looks at waveforms.
+
+### "Offsets develop N months after calibration" — CONFOUNDED, NOT A FINDING
+Tempting, and it looked strong:
+
+| unit | suspect before latest cal | after |
+|---|---|---|
+| BE18438 | 0 / 196 | 87 / 97 |
+| BE12599 | 0 / 62 | 6 / 15 |
+| BE11529 | 0 / 82 | 4 / 17 |
+| BE9558 | 38 / 189 | 0 / 7 |
+
+But bucketing suspects by months-since-calibration gives **one unit per bucket**:
+`0–3mo={BE11529}`, `3–6 & 6–9mo={BE18438}`, `9–12mo={BE9558}`,
+`12–15mo={BE12599}`. The apparent "51% failure rate at 6–9 months" is entirely
+BE18438's single February 2026 episode. Five units with roughly one episode
+each cannot support a population trend. **Do not re-derive this.**
+
+Also note: all affected units are calibrated on a **~12–13 month cadence**, so
+"sent to Instantel" is the routine annual schedule, not evidence of a
+fault-driven return.
+
+---
+
+## 7. Open questions
+
+### Q1 — Is it a latched bad zero or analog degradation?
+The question that decides everything. A latched zero is correctable (possibly
+over the wire); degradation means a repair. Instantel's 2027–2069 rule implies
+*both* populations exist, with the split roughly 10/90 in the field.
+
+**BE12599 is the natural experiment** — faulting as of 2026-08-23. Read its
+values, run the autozero, read them again.
+
+### Q2 — Can we read the autozero numbers over the wire? (best lead)
+Instantel says to check *"the lists of numbers in the **X1 and X8 gains**"* —
+4 sensors × 2 gains = **8 channels**. The protocol reference already documents
+an unimplemented command with exactly that shape:
+
+```
+SUB 0x0E  ->  RSP 0xF1     "channel sensor data"
+  2-step read; channel selector in params[6:8] = 0x0000..0x0007
+  data length 0x0A (10 bytes) per channel
+```
+
+Blastware's *Unit Channel Test* sequence:
+`POLL×N → 0x15 → 0x01 → 0x08 → 0x01 → 0x0E×8 → 0x98×2 → 0x0E×8`
+— note the **second `0x0E` pass carries live ADC readings**.
+
+**Hypothesis (untested):** `0x0E` returns the numbers Instantel wants compared
+against 2027–2069. If true, SFM could diagnose an offset remotely *and* predict
+whether a re-zero will succeed — converting a 10%/90% shipping gamble into a
+decision made before packing a box.
+
+**How to test.** `bridges/ach_mitm.py` is a generic TCP proxy:
+
+```bash
+python bridges/ach_mitm.py --bw-host <MODEM_IP> --bw-port 9034 --listen-port 9999
+```
+
+Point Blastware at the proxy and run **Unit Channel Test**.
+⚠ In this topology the output filenames are reversed — the tool labels the
+*connecting* side "unit", so `raw_s3_*.bin` holds Blastware's bytes and
+`raw_bw_*.bin` the unit's.
+
+Capture priority: (1) BE12599 while faulting, (2) a known-good unit as control,
+(3) before/after an autozero on the same unit. Eight 10-byte payloads with an
+expected value near 2048 is a very constrained puzzle.
+
+### Q3 — What is the mild-offset rate?
+Unmeasurable from event files (§2). Only the A/D-mode check sees it. Would
+need a fleet sweep in A/D mode, or Q2 to succeed.
+
+### Q4 — Does an offset recur on the same unit after service?
+BE9558 shows episodes in 2020 and 2026; BE18438 twice in four months. Suggestive
+of recurrence, but service records aren't in the data — only calibration dates.
+
+---
+
+## 8. Data and tooling
+
+| what | where |
+|---|---|
+| detector | `scratch/offset_scan.py` |
+| current results | `/home/serversdown/dl2-archive/offset_archive.csv` |
+| earlier candidate list (Terra-View DB, 274 events) | `scratch/offset_candidates.csv` |
+| archive working copy | `/home/serversdown/dl2-archive/files/` |
+| archive source | NAS `DeathStar` 10.0.0.2, `/volume1/Uploads/TMI/DL2-Event-backup-8-25-26/Event/autocall home/` |
+
+⚠ **The DL2 export keeps a byte-identical `Sent/` mirror of its root.** 13,077
+waveform paths are 6,577 distinct files. Always dedupe by basename — this
+doubled two reported figures before it was caught.
+
+---
+
+## 9. Chronology
+
+| date | event |
+|---|---|
+| 2026-08-25 | Reported as a *waveform decode bug* — traces with a DC offset. Investigation shows the offset is **real device data**; the decoder is correct. |
+| 2026-08-25 | Brian relays his dad's description: a known hardware fault called an "offset"; usually sent to Instantel. |
+| 2026-08-25 | First detection pass over the Terra-View DB: **2 of 21 units**, 274 events, 11 months. Flagged as vulnerable to survivorship bias — flooded events were routinely deleted. |
+| 2026-08-25 | Condensation hypothesis proposed, then **killed by its own controls**. |
+| 2026-08-25 | Parked pending the multi-year archive. |
+| 2026-08-28 | DL2 archive pulled (33 GB, 546k files; 6.6 GB working set). |
+| 2026-08-28 | Archive scan: **6 of 45 units**, 283 events, 15 episodes. Prior base rate **confirmed**, not overturned. |
+| 2026-08-28 | Clipping ruled out; `m/p` established as the discriminator; BE11007 reclassified as probably a real blast. |
+| 2026-08-28 | Calibration-timing correlation attempted and **rejected as confounded**. |
+| 2026-08-28 | Instantel FAQs supplied: autozero procedure, the **2027–2069** window, the **>5 counts** threshold. Explains the ~10% re-zero success rate. |
+| 2026-08-28 | Bimodality established; sensor check proven **blind** to offsets; `SUB 0x0E` identified as the best open lead. |

minimateplus/histogram_codec.py

@@ -413,10 +413,18 @@ def detect_multi_interval_stride(body: bytes) -> Optional[int]:
 
         if (_ctr(stride) - _ctr(0)) & 0xFFFF != 1:
             continue
-        # confirm on a third block when the body is long enough
-        if 2 * stride + _MULTI_HEADER_LEN <= len(body):
-            if not _is_multi_header(body, 2 * stride):
-                continue
+        # Confirm on a third block WHEN ONE IS ACTUALLY PRESENT.  A body can
+        # be longer than two strides and still hold only two real blocks: a
+        # final *partial* block leaves trailing padding.  E.g. 51 intervals at
+        # 2 s = one full 30-interval block + a 21-interval remainder, in a
+        # 2787-byte body — long enough to demand a third header at 1224 that
+        # does not exist.  Requiring it unconditionally threw away the correct
+        # stride and the file decoded to nothing (BE18193 T193L0XM.CI0H).
+        # The block-counter check above is the decisive anti-false-positive
+        # test; this one is corroboration, so a missing third header means
+        # end-of-stream, not disqualification.
+        if (2 * stride + _MULTI_HEADER_LEN <= len(body)
+                and _is_multi_header(body, 2 * stride)):
             if (_ctr(2 * stride) - _ctr(stride)) & 0xFFFF != 1:
                 continue
         return stride

scratch/offset_scan.py

@@ -0,0 +1,171 @@
+#!/usr/bin/env python3
+"""Scan series-3 waveform binaries for the 'offset' hardware fault.
+
+A healthy geophone trace is centred on zero.  An offset unit sits displaced,
+so the channel mean approaches its own peak.  Detector (unchanged from the
+2026-08-25 run, see memory note `offset-archive-analysis-backlog`):
+
+    dominant-axis  |mean| / peak > 0.7
+    AND            |mean| >= 0.9 * the unit's geo trigger level
+
+Trigger level is read from a paired _ASCII.TXT where one exists, otherwise
+from a per-serial median learned across that unit's ASCII files, otherwise
+--default-trigger.
+
+Serial is decoded from the BW filename: prefix letter encodes thousands
+(chr(ord('B') + n)), next 3 digits the remainder -- T193 -> BE18193.
+
+Usage:
+  python scratch/offset_scan.py --dir <path> [--jobs N] --out offsets.csv
+"""
+from __future__ import annotations
+
+import argparse, csv, json, re, sys
+from collections import defaultdict
+from concurrent.futures import ProcessPoolExecutor, as_completed
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from minimateplus.event_file_io import read_blastware_file
+from minimateplus.bw_ascii_report import parse_report
+
+GEO = ("Tran", "Vert", "Long")
+_GEO_FS_COUNTS = 32000.0
+_WAVE_RE = re.compile(r"\.[A-Za-z0-9]{2}0[Ww]$")
+_STEM_RE = re.compile(r"^([B-Z])(\d{3})")
+
+MEAN_OVER_PEAK_MIN = 0.7
+TRIGGER_FRACTION   = 0.9
+
+
+def serial_from_name(name: str):
+    m = _STEM_RE.match(name)
+    if not m:
+        return None
+    letter, digits = m.group(1), m.group(2)
+    return f"BE{(ord(letter) - ord('B')) * 1000 + int(digits)}"
+
+
+def counts_to_ips(c, gr):
+    return c * (gr or 10.0) / _GEO_FS_COUNTS
+
+
+def scan_one(path_str: str, default_trigger: float) -> dict | None:
+    p = Path(path_str)
+    try:
+        gr, trig = 10.0, None
+        ap = p.with_name(p.name.replace(".", "_", 1) + "_ASCII.TXT") \
+            if False else p.parent / (p.stem + "_" + p.suffix.lstrip(".") + "_ASCII.TXT")
+        if ap.exists():
+            rep = parse_report(ap.read_text(errors="replace"))
+            gr = rep.geo_range_ips or 10.0
+            trig = rep.geo_trigger_level_ips
+        ev = read_blastware_file(p)
+        s = ev.raw_samples or {}
+        if not all(s.get(c) for c in GEO):
+            return None
+        best = None
+        for ch in GEO:
+            arr = s[ch]
+            n = len(arr)
+            if n == 0:
+                continue
+            mean = sum(arr) / n
+            peak = max(abs(v) for v in arr)
+            if peak == 0:
+                continue
+            ratio = abs(mean) / peak
+            if best is None or peak > best["peak_counts"]:
+                best = {"channel": ch, "mean_counts": mean,
+                        "peak_counts": peak, "ratio": ratio}
+        if best is None:
+            return None
+        ts = ev.timestamp
+        return {
+            "serial": serial_from_name(p.name) or "?",
+            "timestamp": (f"{ts.year:04d}-{ts.month:02d}-{ts.day:02d}T"
+                          f"{ts.hour:02d}:{ts.minute:02d}:{ts.second:02d}") if ts else "",
+            "filename": p.name,
+            "channel": best["channel"],
+            "offset_ips": round(counts_to_ips(best["mean_counts"], gr), 4),
+            "peak_ips": round(counts_to_ips(best["peak_counts"], gr), 4),
+            "mean_over_peak": round(best["ratio"], 3),
+            "trigger_level_ips": trig if trig is not None else "",
+            "geo_range_ips": gr,
+        }
+    except Exception:
+        return None
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--dir", required=True)
+    ap.add_argument("--jobs", type=int, default=4)
+    ap.add_argument("--limit", type=int, default=0)
+    ap.add_argument("--default-trigger", type=float, default=0.2)
+    ap.add_argument("--out", required=True)
+    a = ap.parse_args()
+
+    # The DL2 export keeps a byte-identical `Sent/` mirror of the root, so
+    # enumerate paths but keep only the first occurrence of each basename —
+    # otherwise every event is counted twice.
+    seen = set()
+    files = []
+    for q in sorted(Path(a.dir).rglob("*")):
+        if q.is_file() and _WAVE_RE.search(q.name) and q.name not in seen:
+            seen.add(q.name)
+            files.append(q)
+    if a.limit:
+        files = files[: a.limit]
+    print(f"waveform binaries to scan: {len(files)}", flush=True)
+
+    rows = []
+    with ProcessPoolExecutor(max_workers=a.jobs) as ex:
+        futs = [ex.submit(scan_one, str(p), a.default_trigger) for p in files]
+        for n, f in enumerate(as_completed(futs), 1):
+            r = f.result()
+            if r:
+                rows.append(r)
+            if n % 2000 == 0:
+                print(f"  {n}/{len(files)}", flush=True)
+
+    # learn per-serial trigger levels from the rows that had an ASCII
+    by_serial = defaultdict(list)
+    for r in rows:
+        if r["trigger_level_ips"] != "":
+            by_serial[r["serial"]].append(float(r["trigger_level_ips"]))
+    med = {}
+    for k, v in by_serial.items():
+        v.sort()
+        med[k] = v[len(v) // 2]
+
+    for r in rows:
+        if r["trigger_level_ips"] == "":
+            r["trigger_level_ips"] = med.get(r["serial"], a.default_trigger)
+        r["suspect"] = int(
+            r["mean_over_peak"] > MEAN_OVER_PEAK_MIN
+            and abs(r["offset_ips"]) >= TRIGGER_FRACTION * float(r["trigger_level_ips"])
+        )
+
+    cols = ["serial", "timestamp", "filename", "channel", "offset_ips", "peak_ips",
+            "mean_over_peak", "trigger_level_ips", "geo_range_ips", "suspect"]
+    with open(a.out, "w", newline="") as fh:
+        w = csv.DictWriter(fh, fieldnames=cols)
+        w.writeheader()
+        w.writerows(rows)
+
+    sus = [r for r in rows if r["suspect"]]
+    print(f"\nscanned {len(rows)} decodable waveforms")
+    print(f"suspect events: {len(sus)}")
+    per = defaultdict(int)
+    for r in sus:
+        per[r["serial"]] += 1
+    print(f"units with >=1 suspect event: {len(per)} of {len({r['serial'] for r in rows})}")
+    for s, n in sorted(per.items(), key=lambda x: -x[1])[:20]:
+        print(f"  {s:10} {n}")
+    print(f"\nwrote {a.out}")
+
+
+if __name__ == "__main__":
+    main()

scratch/verify_against_ascii.py

@@ -0,0 +1,221 @@
+#!/usr/bin/env python3
+"""Verify the series-3 decoder against preserved Blastware ASCII exports.
+
+Pairs each `<stem>_<ext>_ASCII.TXT` with its binary `<stem>.<ext>`, decodes the
+binary with the production codec, and compares against BW's own export:
+
+  waveform  — per-channel sample counts, then every sample value
+  histogram — interval count, then every per-interval channel peak
+
+ADC counts convert as  ips = counts * geo_range_ips / 32000  (1 decoder unit =
+16 counts = 0.005 in/s at the 10 in/s range; see CLAUDE.md).
+
+Usage:
+  python scratch/verify_against_ascii.py --dir <path> [--limit N] [--jobs N]
+                                         [--out results.json] [--kind w|h|all]
+"""
+from __future__ import annotations
+
+import argparse, json, re, sys, traceback
+from concurrent.futures import ProcessPoolExecutor, as_completed
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from minimateplus.event_file_io import read_blastware_file
+from minimateplus.bw_ascii_report import parse_report_file, parse_report
+
+GEO = ("Tran", "Vert", "Long")
+_GEO_FS_COUNTS = 32000.0
+_ASCII_SUFFIX_RE = re.compile(r"_ASCII\.TXT$", re.IGNORECASE)
+
+
+def binary_for(ascii_path: Path) -> Path:
+    """H907KXOW_WC0H_ASCII.TXT -> H907KXOW.WC0H"""
+    stem = _ASCII_SUFFIX_RE.sub("", ascii_path.name)
+    if "_" not in stem:
+        return ascii_path.with_name(stem)
+    head, _, ext = stem.rpartition("_")
+    return ascii_path.with_name(f"{head}.{ext}")
+
+
+def counts_to_ips(counts, geo_range_ips):
+    r = geo_range_ips if geo_range_ips else 10.0
+    return counts * r / _GEO_FS_COUNTS
+
+
+def agrees(got, exp, geo_range_ips, tol=0.0006):
+    """True when decoded `got` matches BW's exported `exp`.
+
+    Saturation carve-out: when an event clips, BW clamps its export to the
+    channel's range maximum (and writes OORANGE for the summary PPV), while
+    the decoder faithfully reproduces raw counts that can sit a decoder unit
+    or two past nominal full scale (32016 counts observed = 10.005 in/s on
+    the 10 in/s range).  Same sign and both at/above the ceiling is agreement,
+    not a decode error.
+    """
+    if abs(got - exp) <= tol:
+        return True
+    r = geo_range_ips if geo_range_ips else 10.0
+    if abs(exp) >= r - tol and abs(got) >= r - tol and (got >= 0) == (exp >= 0):
+        return True
+    return False
+
+
+def parse_interval_table(text: str):
+    """Histogram interval rows: time, Tpk, Tfq, Vpk, Vfq, Lpk, Lfq, PVS, ..., micdB, micfq"""
+    rows = []
+    seen_header = False
+    for line in text.splitlines():
+        if "\t" not in line:
+            continue
+        cols = [c.strip().strip('"') for c in line.split("\t")]
+        cols = [c for c in cols if c != ""]
+        if not seen_header:
+            if any(c in ("Tran", "Vert", "Long") for c in cols):
+                seen_header = True
+            continue
+        if len(cols) < 7:
+            continue
+        if not re.match(r"^\d{1,2}:\d{2}:\d{2}$", cols[0]):
+            continue
+        def num(s):
+            try:
+                return float(s)
+            except ValueError:
+                return None
+        rows.append({"time": cols[0], "Tran": num(cols[1]),
+                     "Vert": num(cols[3]), "Long": num(cols[5])})
+    return rows
+
+
+def check_one(ascii_path_str: str) -> dict:
+    ap = Path(ascii_path_str)
+    bp = binary_for(ap)
+    res = {"ascii": ap.name, "binary": bp.name, "status": "?",
+           "kind": None, "detail": ""}
+    try:
+        if not bp.exists():
+            res["status"] = "no_binary"
+            return res
+        text = ap.read_text(errors="replace")
+        rep = parse_report(text, parse_samples=True)
+        ev = read_blastware_file(bp)
+        gr = rep.geo_range_ips
+        res["kind"] = kind = ("histogram"
+                              if (rep.event_type or "").lower().startswith(("full histogram", "histogram"))
+                              else "waveform")
+        samples = ev.raw_samples or {}
+        dec_n = {c: len(samples.get(c) or []) for c in GEO}
+
+        if kind == "histogram":
+            rows = parse_interval_table(text)
+            res["n_ascii"] = len(rows)
+            res["n_decoded"] = dec_n["Tran"]
+            if not rows:
+                res["status"] = "no_ascii_table"
+                return res
+            if dec_n["Tran"] == 0:
+                res["status"] = "decode_empty"
+                return res
+            if dec_n["Tran"] != len(rows):
+                res["status"] = "count_mismatch"
+                res["detail"] = f"decoded {dec_n['Tran']} vs ascii {len(rows)}"
+                return res
+            bad = 0
+            worst = 0.0
+            for i, row in enumerate(rows):
+                for ch in GEO:
+                    exp = row[ch]
+                    if exp is None:
+                        continue
+                    got = counts_to_ips(samples[ch][i], gr)
+                    if not agrees(got, exp, gr):
+                        bad += 1
+                        worst = max(worst, abs(got - exp))
+            res["worst_abs"] = round(worst, 6)
+            res["status"] = "exact" if bad == 0 else "value_mismatch"
+            if bad:
+                res["detail"] = f"{bad} interval-channel values off"
+            return res
+
+        # waveform
+        asc = rep.samples or []
+        res["n_ascii"] = len(asc)
+        res["n_decoded"] = dec_n["Tran"]
+        if not asc:
+            res["status"] = "no_ascii_table"
+            return res
+        if dec_n["Tran"] == 0:
+            res["status"] = "decode_empty"
+            return res
+        if len({dec_n[c] for c in GEO}) != 1:
+            res["status"] = "channel_len_mismatch"
+            res["detail"] = str(dec_n)
+            return res
+        if dec_n["Tran"] != len(asc):
+            res["status"] = "count_mismatch"
+            res["detail"] = f"decoded {dec_n['Tran']} vs ascii {len(asc)}"
+            return res
+        bad = 0
+        worst = 0.0
+        for i, quad in enumerate(asc):
+            for j, ch in enumerate(GEO):
+                exp = quad[j]
+                got = counts_to_ips(samples[ch][i], gr)
+                if not agrees(got, exp, gr):
+                    bad += 1
+                    worst = max(worst, abs(got - exp))
+        res["worst_abs"] = round(worst, 6)
+        res["status"] = "exact" if bad == 0 else "value_mismatch"
+        if bad:
+            res["detail"] = f"{bad} sample values off"
+        return res
+    except Exception as e:
+        res["status"] = "error"
+        res["detail"] = f"{type(e).__name__}: {e}"
+        return res
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--dir", required=True)
+    ap.add_argument("--limit", type=int, default=0)
+    ap.add_argument("--jobs", type=int, default=8)
+    ap.add_argument("--kind", choices=["w", "h", "all"], default="all")
+    ap.add_argument("--out", default=None)
+    a = ap.parse_args()
+
+    root = Path(a.dir)
+    files = sorted(p for p in root.rglob("*")
+                   if p.is_file() and p.name.upper().endswith("_ASCII.TXT"))
+    if a.kind != "all":
+        want = "0W" if a.kind == "w" else "0H"
+        files = [p for p in files
+                 if _ASCII_SUFFIX_RE.sub("", p.name).upper().endswith(want)]
+    if a.limit:
+        files = files[: a.limit]
+    print(f"pairs to check: {len(files)}", flush=True)
+
+    out = []
+    from collections import Counter
+    tally = Counter()
+    with ProcessPoolExecutor(max_workers=a.jobs) as ex:
+        futs = {ex.submit(check_one, str(p)): p for p in files}
+        for n, f in enumerate(as_completed(futs), 1):
+            r = f.result()
+            out.append(r)
+            tally[(r["kind"], r["status"])] += 1
+            if n % 500 == 0:
+                print(f"  {n}/{len(files)}", flush=True)
+
+    print("\n=== results ===")
+    for (kind, status), n in sorted(tally.items(), key=lambda x: -x[1]):
+        print(f"  {str(kind):10} {status:22} {n}")
+    if a.out:
+        Path(a.out).write_text(json.dumps(out, indent=1))
+        print(f"\nwrote {a.out}")
+
+
+if __name__ == "__main__":
+    main()

sfm/database.py

@@ -603,56 +603,107 @@ class SeismoDb:
             ).fetchall()
         return [dict(r) for r in rows]
 
-    def find_twins(self, event_id: str, *, window_seconds: int = 300) -> list[dict]:
+    def find_twins(self, event_id: str, *, window_seconds: int | None = None) -> list[dict]:
         """
-        Find this event's histogram/waveform twin(s): rows sharing the same
-        serial and an identical peak_vector_sum, whose timestamp falls
-        within ``window_seconds`` of this event's timestamp.  Excludes the
-        event itself.  Returns [] if the event or any required field
-        (serial / peak_vector_sum / timestamp) is missing.
+        Find this event's histogram/waveform twin(s): the SAME physical event
+        recorded both as a scheduled histogram and as a triggered waveform.
 
-        Caveat: identical-PVS matching is a proxy for "same physical event
-        recorded twice," not a guarantee. In the rare case where the device
-        clamps/saturates PVS (clamped to sqrt(3) * geo_range), two distinct
-        saturated events on the same serial within the window can share the
-        same clamped PVS value and be matched as twins even though they are
-        different events. This is harmless in practice — false_trigger/
-        reviewed_real are derived/index columns re-derivable from the
-        sidecar source of truth — but worth knowing if twin counts look
-        surprising on a saturated/clamped run.
+        A real trigger is captured twice — once as a triggered waveform (stamped
+        at the trigger instant) and once inside the scheduled histogram whose
+        interval contains it (stamped at the histogram's interval start, e.g. the
+        7am/7pm call-in). The two can be HOURS apart in time yet report the same
+        serial and identical peak_vector_sum. Twins are therefore matched by:
+
+          * same serial,
+          * identical peak_vector_sum,
+          * OPPOSITE record type (one histogram, one waveform), and
+          * the waveform's timestamp falls within the histogram's interval —
+            from a histogram's timestamp up to the next histogram (same serial).
+
+        This replaces the old ±``window_seconds`` heuristic, which silently
+        missed twins more than a few minutes apart (a histogram's interval-start
+        stamp and the trigger instant routinely differ by hours). ``window_seconds``
+        is still accepted for backward compatibility but is ignored.
+
+        Returns [] if the event or a required field (serial / peak_vector_sum /
+        timestamp) is missing.
+
+        Caveat: identical-PVS matching remains a proxy for "same physical event"
+        — if the device clamps/saturates PVS (to sqrt(3) * geo_range), two
+        distinct saturated events could share a PVS. The added opposite-type and
+        interval constraints make a false pairing far less likely than the old
+        time-window match, and false_trigger/reviewed_real stay re-derivable from
+        the sidecar source of truth.
         """
+        def _parse(ts):
+            if not ts:
+                return None
+            try:
+                return datetime.datetime.fromisoformat(str(ts).replace(" ", "T"))
+            except ValueError:
+                return None
+
+        def _is_hist(rt):
+            return str(rt or "").lower().startswith("hist")
+
         row = self.get_event(event_id)
         if not row:
             return []
-        serial = row.get("serial"); pvs = row.get("peak_vector_sum"); ts = row.get("timestamp")
-        if serial is None or pvs is None or not ts:
+        serial = row.get("serial"); pvs = row.get("peak_vector_sum")
+        t_target = _parse(row.get("timestamp"))
+        if serial is None or pvs is None or t_target is None:
             return []
-        try:
-            t = datetime.datetime.fromisoformat(ts.replace(" ", "T"))
-        except ValueError:
-            return []
-        lo = (t - datetime.timedelta(seconds=window_seconds)).isoformat()
-        hi = (t + datetime.timedelta(seconds=window_seconds)).isoformat()
-        with self._connect() as conn:
-            rows = conn.execute(
-                "SELECT * FROM events WHERE serial=? AND id!=? AND peak_vector_sum=? "
-                "AND timestamp BETWEEN ? AND ?",
-                (serial, event_id, pvs, lo, hi),
-            ).fetchall()
-        return [dict(r) for r in rows]
+        target_hist = _is_hist(row.get("record_type"))
 
-    def propagate_review_to_twins(self, event_id: str, *, window_seconds: int = 300) -> list[str]:
+        with self._connect() as conn:
+            cand_rows = [dict(r) for r in conn.execute(
+                "SELECT * FROM events WHERE serial=? AND id!=? AND peak_vector_sum=?",
+                (serial, event_id, pvs)).fetchall()]
+            hist_ts = [r["timestamp"] for r in conn.execute(
+                "SELECT timestamp FROM events WHERE serial=? AND lower(record_type) LIKE 'hist%'",
+                (serial,)).fetchall()]
+
+        # Histogram interval-start times for this serial, sorted, to bound intervals.
+        starts = sorted(x for x in (_parse(t) for t in hist_ts) if x is not None)
+
+        def _interval_end(h_start):
+            # The next histogram strictly after h_start bounds the interval; else open-ended.
+            for x in starts:
+                if x > h_start:
+                    return x
+            return None
+
+        def _covers(h_start, w_time):
+            end = _interval_end(h_start)
+            return h_start <= w_time and (end is None or w_time < end)
+
+        twins = []
+        for c in cand_rows:
+            if _is_hist(c.get("record_type")) == target_hist:
+                continue  # twins are strictly cross-type (one histogram, one waveform)
+            c_time = _parse(c.get("timestamp"))
+            if c_time is None:
+                continue
+            h_start, w_time = (t_target, c_time) if target_hist else (c_time, t_target)
+            if _covers(h_start, w_time):
+                twins.append(c)
+        return twins
+
+    def propagate_review_to_twins(self, event_id: str, *, window_seconds: int | None = None) -> list[str]:
         """
         Copy this event's `false_trigger`/`reviewed_real` columns onto each
         of its histogram/waveform twins (see `find_twins`), so flagging one
         twin flags both.  Returns the list of twin ids updated.
+
+        ``window_seconds`` is accepted for backward compatibility but ignored;
+        twin matching is now interval-based (see `find_twins`).
         """
         row = self.get_event(event_id)
         if not row:
             return []
         ft = 1 if row.get("false_trigger") else 0
         real = 1 if row.get("reviewed_real") else 0
-        twins = self.find_twins(event_id, window_seconds=window_seconds)
+        twins = self.find_twins(event_id)
         moved = []
         with self._connect() as conn:
             for tw in twins:

tests/test_find_twins.py

@@ -1,33 +1,71 @@
-import datetime
+import sqlite3
 from sfm.database import SeismoDb
 from minimateplus.models import Event, Timestamp
 
 
-def _ins(db, key, serial, pvs, ts):
+def _ins(db, key, serial, pvs, ts, record_type="Waveform"):
     ev = Event(index=0)
     ev._waveform_key = bytes.fromhex(key)
     ev.timestamp = ts
-    # peak_vector_sum comes from peak_values; simplest: insert then UPDATE pvs directly
     db.insert_events([ev], serial=serial)
     row = [r for r in db.query_events(serial=serial) if r["waveform_key"] == key][0]
-    import sqlite3
     with sqlite3.connect(db.db_path) as c:
-        c.execute("UPDATE events SET peak_vector_sum=? WHERE id=?", (pvs, row["id"]))
+        c.execute("UPDATE events SET peak_vector_sum=?, record_type=? WHERE id=?",
+                  (pvs, record_type, row["id"]))
     return row["id"]
 
 
-def test_find_twins_matches_same_serial_pvs_near_time(tmp_path):
+def _ts(hour, minute, second=0, day=25):
+    return Timestamp(raw=b"", flag=0x10, year=2026, unknown_byte=0, month=2, day=day,
+                     hour=hour, minute=minute, second=second)
+
+
+def test_histogram_and_waveform_twin_across_hours(tmp_path):
+    # The real UM12947 case: histogram stamped at its 7pm interval start, the
+    # triggered waveform 75 min later — same serial + identical PVS. The old
+    # ±5-min window missed this; interval matching catches it, both directions.
     db = SeismoDb(tmp_path / "s.db")
-    base = Timestamp(raw=b"", flag=0x10, year=2026, unknown_byte=0, month=2, day=25, hour=20, minute=19, second=5)
-    twin = Timestamp(raw=b"", flag=0x10, year=2026, unknown_byte=0, month=2, day=25, hour=20, minute=19, second=45)
-    far = Timestamp(raw=b"", flag=0x10, year=2026, unknown_byte=0, month=2, day=25, hour=21, minute=0, second=0)
-    # d needs a timestamp distinct from `twin` (UNIQUE(serial, timestamp) would
-    # otherwise collide with b and UPSERT onto its row instead of inserting a
-    # new one) while staying near `base` in time.
-    near = Timestamp(raw=b"", flag=0x10, year=2026, unknown_byte=0, month=2, day=25, hour=20, minute=19, second=44)
-    a = _ins(db, "01110001", "BE1", 0.4763, base)
-    b = _ins(db, "01110002", "BE1", 0.4763, twin)   # twin: same serial+pvs, 40s apart
-    c = _ins(db, "01110003", "BE1", 0.4763, far)    # same pvs but >window away
-    d = _ins(db, "01110004", "BE1", 0.9999, near)   # near time but different pvs
-    ids = {r["id"] for r in db.find_twins(a, window_seconds=300)}
-    assert ids == {b}
+    hist_pm = _ins(db, "01110001", "BE1", 0.4763, _ts(19, 31, 17), "Histogram")
+    wave    = _ins(db, "01110002", "BE1", 0.4763, _ts(20, 46, 44), "Waveform")
+    _ins(db, "01110003", "BE1", 0.0100, _ts(7, 0, 0, day=26), "Histogram")  # bounds the interval
+    assert {r["id"] for r in db.find_twins(hist_pm)} == {wave}
+    assert {r["id"] for r in db.find_twins(wave)} == {hist_pm}
+
+
+def test_same_type_not_twinned(tmp_path):
+    # Two waveforms, same serial + PVS, seconds apart → NOT twins (cross-type only).
+    db = SeismoDb(tmp_path / "s.db")
+    a = _ins(db, "01110001", "BE1", 0.4763, _ts(20, 19, 5), "Waveform")
+    _ins(db, "01110002", "BE1", 0.4763, _ts(20, 19, 45), "Waveform")
+    assert db.find_twins(a) == []
+
+
+def test_waveform_matches_only_the_containing_interval(tmp_path):
+    # Two overnight intervals with the same PVS; a waveform in the SECOND interval
+    # must twin with that histogram, never the first — even though PVS matches both.
+    db = SeismoDb(tmp_path / "s.db")
+    h1 = _ins(db, "01110001", "BE1", 0.4763, _ts(19, 0, 0, day=25), "Histogram")
+    h2 = _ins(db, "01110002", "BE1", 0.4763, _ts(7, 0, 0, day=26), "Histogram")
+    w = _ins(db, "01110003", "BE1", 0.4763, _ts(8, 0, 0, day=26), "Waveform")
+    assert {r["id"] for r in db.find_twins(w)} == {h2}
+    assert w not in {r["id"] for r in db.find_twins(h1)}
+
+
+def test_different_pvs_not_twinned(tmp_path):
+    db = SeismoDb(tmp_path / "s.db")
+    h = _ins(db, "01110001", "BE1", 0.4763, _ts(19, 0, 0), "Histogram")
+    _ins(db, "01110002", "BE1", 0.9999, _ts(20, 0, 0), "Waveform")  # different PVS
+    assert db.find_twins(h) == []
+
+
+def test_open_ended_latest_interval(tmp_path):
+    # A waveform after the latest histogram (nothing bounds the interval) still twins.
+    db = SeismoDb(tmp_path / "s.db")
+    h = _ins(db, "01110001", "BE1", 0.4763, _ts(19, 0, 0), "Histogram")
+    w = _ins(db, "01110002", "BE1", 0.4763, _ts(23, 30, 0), "Waveform")
+    assert {r["id"] for r in db.find_twins(h)} == {w}
+
+
+def test_missing_fields_returns_empty(tmp_path):
+    db = SeismoDb(tmp_path / "s.db")
+    assert db.find_twins("nonexistent-id") == []

tests/test_histogram_codec.py

@@ -643,3 +643,38 @@ def test_multi_interval_matches_blastware_ascii_exactly():
                 assert hz is None
             elif not cell.startswith("<"):
                 assert hz is not None and abs(hz - float(cell)) <= max(0.55, float(cell) * 0.02)
+
+
+def test_partial_final_block_is_not_disqualified_by_missing_third_header():
+    """A body can exceed two strides yet hold only two real blocks.
+
+    Regression for BE18193 `T193L0XM.CI0H` — 51 intervals at 2 s = one full
+    30-interval block plus a 21-interval remainder, in a body long enough to
+    demand a third block header at ``2 * stride`` that does not exist.  The
+    third-block confirmation used to be mandatory whenever the body was long
+    enough, so the correct stride was discarded and the file decoded to
+    nothing.  A missing third header means end-of-stream, not disqualification;
+    the block-counter check is the decisive anti-false-positive test.
+    """
+    full = [(1, 1, 2, 2, 3, 3, 4, 4)] * 30
+    partial = [(5, 5, 6, 6, 7, 7, 8, 8)] * 21
+    body = (_mk_multi_block(full, ctr=256)
+            + _mk_multi_block(partial, ctr=257)
+            + b"\xff" * 700)      # trailing padding past 2 * stride
+    stride = 12 + 20 * 30
+    assert 2 * stride + 6 <= len(body), "padding must reach past two strides"
+    # the whole point: a third header is absent, and that must not disqualify
+    assert detect_multi_interval_stride(body) == stride
+    recs = walk_multi_interval_blocks(body)
+    assert len(recs) == 51
+    assert recs[0]["t_peak"] == 1
+    assert recs[-1]["t_peak"] == 5
+
+
+def test_third_block_still_rejects_a_mismatched_counter():
+    """The corroboration must still bite when a third block IS present."""
+    ivs = [(1, 1, 2, 2, 3, 3, 4, 4)] * 4
+    body = (_mk_multi_block(ivs, ctr=256)
+            + _mk_multi_block(ivs, ctr=257)
+            + _mk_multi_block(ivs, ctr=999))   # counter jumps — not consecutive
+    assert detect_multi_interval_stride(body) != 12 + 20 * 4

tests/test_twin_propagation.py

@@ -3,31 +3,34 @@ from sfm.database import SeismoDb
 from minimateplus.models import Event, Timestamp
 
 
-def _ins(db, key, serial, pvs, ts):
+def _ins(db, key, serial, pvs, ts, record_type="Waveform"):
     ev = Event(index=0)
     ev._waveform_key = bytes.fromhex(key)
     ev.timestamp = ts
-    # peak_vector_sum comes from peak_values; simplest: insert then UPDATE pvs directly
     db.insert_events([ev], serial=serial)
     row = [r for r in db.query_events(serial=serial) if r["waveform_key"] == key][0]
     with sqlite3.connect(db.db_path) as c:
-        c.execute("UPDATE events SET peak_vector_sum=? WHERE id=?", (pvs, row["id"]))
+        c.execute("UPDATE events SET peak_vector_sum=?, record_type=? WHERE id=?",
+                  (pvs, record_type, row["id"]))
     return row["id"]
 
 
-def test_propagate_copies_flags_to_twins(tmp_path):
+def _ts(hour, minute, second=0, day=25):
+    return Timestamp(raw=b"", flag=0x10, year=2026, unknown_byte=0, month=2, day=day,
+                     hour=hour, minute=minute, second=second)
+
+
+def test_propagate_copies_flags_across_hours_apart_twins(tmp_path):
+    # Flagging the waveform FT propagates to its histogram twin 75 min earlier
+    # (the interval matcher pairs them; the old ±5-min window would have missed it).
     db = SeismoDb(tmp_path / "s.db")
-    base = Timestamp(raw=b"", flag=0x10, year=2026, unknown_byte=0, month=2, day=25, hour=20, minute=19, second=5)
-    twin = Timestamp(raw=b"", flag=0x10, year=2026, unknown_byte=0, month=2, day=25, hour=20, minute=19, second=45)
-    other = Timestamp(raw=b"", flag=0x10, year=2026, unknown_byte=0, month=2, day=25, hour=20, minute=19, second=44)
+    hist  = _ins(db, "01110001", "BE1", 0.4763, _ts(19, 31, 17), "Histogram")
+    wave  = _ins(db, "01110002", "BE1", 0.4763, _ts(20, 46, 44), "Waveform")   # twin, 75 min later
+    other = _ins(db, "01110003", "BE1", 0.9999, _ts(20, 20, 0),  "Waveform")   # different pvs
 
-    primary_id = _ins(db, "01110001", "BE1", 0.4763, base)
-    twin_id = _ins(db, "01110002", "BE1", 0.4763, twin)        # twin: same serial+pvs, 40s apart
-    non_twin_id = _ins(db, "01110003", "BE1", 0.9999, other)   # near time but different pvs
+    db.update_event_review(wave, {"false_trigger": True})
+    moved = db.propagate_review_to_twins(wave)
 
-    db.update_event_review(primary_id, {"false_trigger": True})
-    moved = db.propagate_review_to_twins(primary_id)
-
-    assert twin_id in moved
-    assert db.get_event(twin_id)["false_trigger"] == 1
-    assert db.get_event(non_twin_id)["false_trigger"] == 0
+    assert hist in moved
+    assert db.get_event(hist)["false_trigger"] == 1
+    assert db.get_event(other)["false_trigger"] == 0