fix(codec): 40 NN int16 blocks are not capped at NN=8

data_block_len() rejected any `40 NN` block with NN > 0x08. That guard had
no evidence behind it: every corpus available when it was written used only
NN in {1,2,3,4,8}, so it was never exercised. Loud UM12947 events use NN of
12, 16, 20 ... up to 196.

Because walk_body/run stop at the first unrecognised tag rather than
raising, rejecting those blocks surfaced as silently short channels -- e.g.
Tran 1812 / Vert 2132 / Long 2324 on a file whose export carries 2324 for
all three. The real bound is the buffer; the caller additionally clamps to
the record end.

Verified against Thor's own CSV exports for UM12947 (2025-07-14 .. 09-25,
167 waveforms, supplied as CSV.zip):

  length mismatches   22 -> 0
  per-sample exact    1,476,242 / 1,476,249

These are NOT truncated recordings, which was the competing hypothesis --
the exports carry the full sample count.

tests/test_waveform_codec.py asserted the cap as intended behaviour. That
assertion encoded an assumption, not a verified fact, and is replaced with
one pinning the opposite plus the evidence.

Across all three ground-truth corpora: 459 waveform files,
3,807,158 / 3,807,165 samples exact. Production IDFW is now 575/575 with
zero truncations and zero decode failures (median PPV error -0.0007% across
8 units). Series-3 re-verified unchanged at 14,338/14,338.

The 7 residual samples each differ by one 4th-decimal tick and are Thor's
own rounding: intersecting the per-sample rounding constraints over that
corpus is infeasible (binding pair contradict by 2.3e-11, 7e-5 relative),
so no single linear LSB reproduces every printed value. _GEO_LSB_IPS is
already pinned to ~1e-11; do not retune it to chase these.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ru8Lg9HkkYvX9VWWo65SmL

CLAUDE.md

@@ -47,11 +47,20 @@ Read this first when picking the project back up.
   was invisible and Vert came up exactly 512 short), and the histogram
   interval record is **56 bytes, not 72** — `16 × n_channels + 8`, derived per
   segment from the cumulative interval counter, never assumed.
-  **Still open:** 23/575 prod IDFW files (4%), all UM12947 2025-07-14..09-23,
-  stop the walker on tag `40 0c` — `data_block_len()` caps `40 NN` at NN>8 but
-  these use up to 196.  **Diagnosed, NOT verified** — no Thor export exists for
-  those dates.  Do not ship the cap change without one; see
-  `docs/idf_protocol_reference.md`.
+- **`40 NN` blocks are not capped at NN=8 (2026-09-11).**  `data_block_len()`
+  rejected `NN > 0x08`, a guard with no evidence behind it — the corpora
+  available when it was written only used NN ∈ {1,2,3,4,8}.  Loud UM12947
+  events use NN up to 196, and since the walker stops at the first
+  unrecognised tag rather than raising, this surfaced as silently short
+  channels.  Verified on 167 UM12947 waveforms: length mismatches 22 → 0,
+  1,476,242/1,476,249 samples exact.
+- **Production IDFW is now 575/575** — zero truncations, zero decode
+  failures, median PPV error −0.0007% across 8 units (was 41 truncated + 1
+  failing, −3.3%).  Across all three ground-truth corpora: **459 files,
+  3,807,158/3,807,165 samples exact**; the 7 stragglers differ by one
+  4th-decimal tick and are Thor's own rounding — no single linear LSB can
+  reproduce every printed value (the constraints are infeasible by 7e-5
+  relative), so do NOT retune `_GEO_LSB_IPS`.
 - **Open, not blocking:** 14 sensitive-range files show an exact 8x
   (= 10.0/1.25) units discrepancy; `scripts/backfill_sidecars.py --force` also
   inserts DB rows for store files that have none (one-time per store) and the