diff --git a/CLAUDE.md b/CLAUDE.md
index b2ceb14..bf05b6f 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -171,25 +171,39 @@ the `b'STRT'` magic bytes:
 ```
 +0..3   b'STRT'   magic
 +4..5   flags     0xFF 0xFE (single-shot) or 0xFF 0xFD (continuous)
-+6..9   key4      4-byte event key
-+10..13 prev_key4
-+14..15 uint16 BE total_samples  (full event sample-set count) ← confirmed 4-9-26
-+16..17 uint16 BE pretrig_samples (pre-trigger sample-set count)
++6..9   next_key4  ← key of the NEXT stored event (NOT the current event) ← confirmed 2026-04-14
++10..13 prev_key4  ← key of the PREVIOUS stored event                     ← confirmed 2026-04-14
++14..15 UNKNOWN   (values seen: 0xDA63=55907, 0xF38F=62351, 0x5685=22149) — NOT total_samples
++16..17 UNKNOWN   (values seen: 0x0122=290, 0x011A=282, 0x00FA=250)       — NOT pretrig_samples
 +18     uint8     record-MODE byte — NOT rectime in seconds
-+19..20 typically 0x00 0x00
++19..20 0x00 0x00
 ```
 
+**CONFIRMED field values (2026-04-14) from 3 desk-thump events, firmware S338.17:**
+
+| Field | What it is | Confirmed |
+|---|---|---|
+| +4..5 | 0xFFFE single-shot / 0xFFFD continuous | ✅ |
+| +6..9 | next_event_key (NOT current) | ✅ 3 events |
+| +10..13 | prev_event_key | ✅ 3 events |
+| +18 | mode byte: 0x46 ('F') = single-shot, 0x0E = continuous | ✅ |
+
+**UNCONFIRMED — total_samples and pretrig_samples locations unknown:**
+The prior documented offsets (+14..15 for total_samples, +16..17 for pretrig_samples) were
+WRONG — confirmed by cross-checking STRT-derived rectime against compliance record_time
+(4-14-26): all 3 events give STRT-derived rectime of 21–61 s vs actual 3.0 s (ratio 7–20×).
+The "confirmed 4-9-26" note in prior versions was incorrect.
+
+The true offsets for total_samples and pretrig_samples within STRT have not been located.
+**Until they are found, `_decode_a5_waveform` relies on the compliance-config cross-check
+fallback for all total_samples and pretrig_samples values.**
+
 **CRITICAL — strt[18] is a record-mode byte, NOT rectime_seconds (confirmed 2026-04-14):**
 Analysis of 15 distinct STRT records across the 4-9-26 ACH capture shows:
 - `flags=0xFFFE` (single-shot) → `strt[18] = 0x46` ('F') for EVERY event regardless of duration
 - `flags=0xFFFD` (continuous)  → `strt[18] = 0x0E` for EVERY event regardless of duration
 
-The actual record duration (post-trigger seconds) must be computed as:
-```python
-rectime_seconds = int(round((total_samples - pretrig_samples) / sample_rate))
-```
-`_decode_a5_waveform` uses `sample_rate=1024` as a default; the server overrides with
-`compliance_config.sample_rate` when available.  Do NOT use `strt[18]` for rectime.
+Do NOT use `strt[18]` for rectime.
 
 **Pre-trigger time is separate from record_time (confirmed 2026-04-14):**
 Blastware documentation states: "The default Time Scale is -0.25 second to 1 second — this
@@ -197,17 +211,17 @@ negative number accounts for the pre-trigger set for compliance monitoring."  Th
 - `record_time` (3.0 s) is POST-TRIGGER duration only
 - Pre-trigger = 0.25 s = 256 samples at 1024 sps (compliance monitoring standard default)
 - The pre-trigger field has NOT yet been located in the raw compliance config bytes
-- When STRT layout is invalid, `_decode_a5_waveform` falls back to pretrig = 0.25 × sr
+- `_decode_a5_waveform` falls back to pretrig = 0.25 × sr from compliance standard
 - TODO: locate pretrig_time offset in ComplianceConfig — search around anchor or channel blocks
 
 The device bulk-streams zero-padded frames BEYOND the configured record window.  The
 viewer clips `displayCount = total_samples = pretrig + post_trig` to exclude this padding.
 
-**Sanity check — pretrig_samples must be less than total_samples:**
-If `pretrig_samples >= total_samples` the STRT parse is invalid.  Possible causes:
-DLE-stuffed `0x10` byte within prev_key4 or key4 shifted field offsets, or a different
-STRT record variant.  `_decode_a5_waveform` logs `raw strt[0:21]` at WARNING level and
-clamps `pretrig_samples = 0` so the viewer renders (showing the full waveform from t=0).
+**Validity checks in `_decode_a5_waveform`:**
+Check 1: `pretrig_samples >= total_samples` → invalid (original check).
+Check 2: STRT-derived rectime differs from `compliance_config.record_time` by more than 2×
+→ invalid.  Both failures fall back to the compliance-config derived values.
+`_decode_a5_waveform` logs `raw strt[0:21]` at WARNING level on any failure.
 Observed once (2026-04-14) with `strt[16:18] = 0x41 0x01` → pretrig=16641 (impossible).
 Root cause not yet identified — capture the warning log hex dump to diagnose.
 
diff --git a/minimateplus/client.py b/minimateplus/client.py
index e7637fe..afc5245 100644
--- a/minimateplus/client.py
+++ b/minimateplus/client.py
@@ -1389,11 +1389,18 @@ def _decode_a5_waveform(
     #
     # NOTE: strt[8:10] is the LOWER 2 bytes of key4, NOT total_samples.
     # Confirmed from raw_rx capture (4-9-26): strt[14:16] = total_samples. ✅
-    strt = w0[strt_pos : strt_pos + 21]
+    # Extract 32 bytes so we can see past the first 21 — the true total_samples
+    # and pretrig_samples offsets have not yet been confirmed and may be > +20.
+    strt = w0[strt_pos : strt_pos + 32]
     if len(strt) < 21:
         log.warning("_decode_a5_waveform: STRT record truncated (%dB)", len(strt))
         return
 
+    log.info(
+        "_decode_a5_waveform: STRT raw[0:32]: %s",
+        strt[:32].hex(' '),
+    )
+
     total_samples   = struct.unpack_from(">H", strt, 14)[0]
     pretrig_samples = struct.unpack_from(">H", strt, 16)[0]