docs(tide-table): hammer v1 ear-findings measured + v2 rebuild log

PLN re-listened to the forged hammer kit: vox cuts weren't voice-only,
06_vox_breakitdown didn't loop, kit felt below the track's stature. All three
confirmed by measurement and fixed in a v2 rebuild (catch report.md has the
full story): demucs-stem vox bleed sat just −2.4 dB under the cut RMS on the
worst file; the energy-corrected cut bounds were themselves bleed-biased
(06 started a full second before the word); the rotated loop exports had
never been promoted. v2 = roformer-vocals re-cuts on true onsets, a
roformer-instrumental→demucs cascade for ghost-free instrument loops, the
missing 2-bar breakitdown loop, and the rotation finally shipped.

armada/tide-table/performance_notes.md

@@ -197,3 +197,23 @@ loop can be rotated freely — the new seam is an interior continuity point →
 pick rotation with strongest attack at 0); vocals via engine2 BS-Roformer (keeps low
 end far better than htdemucs) then re-cut the same phrase bounds; bass_deep → optional
 harmonic-saturation "audible" variant, or accept as sub layer + in-session shape.
+
+**Hammer kit re-listen (2026-07-19, PLN).** Ear-report on the v1 `hammer` kit:
+(1) "some cases it's not even just the voices" (vox cuts carry instrumental);
+(2) "06_vox_breakitdown doesn't loop well"; (3) overall "we didn't do honor to
+hammer". superfreak folder judged good on re-listen.
+
+All three measured + fixed (v2 rebuild, see catch `report.md` + `review_v2_bleed_metric.py`):
+- Vox bleed confirmed: v1 cuts (demucs stem) held riff energy at only −2.4 dB
+  (03) / −6.5 dB (06) below cut RMS during vocal-silence frames. v2 re-cut from
+  the BS-Roformer vocals → bleed −27..−62 dB below signal.
+- **New lesson beyond the 07-11 note: the CUT BOUNDS were bleed-biased too.** The
+  energy-corrected bounds fitted on the bleedy demucs stem started 06 a full
+  second before the word ("break" is at 202.81, cut at 201.70) and 04 450 ms
+  early. "Re-cut the same phrase bounds" (07-11 fix direction) was insufficient —
+  bounds must be re-anchored on the cleanest vocal render's true onsets.
+- 06 "doesn't loop": it was an 8.30-beat one-shot; v2 adds 09_vox_breakitdown_loop
+  (2.000 bars exact, x4 join 0.00x) and keeps 06 as a full-tail one-shot (4.95 s).
+- Rotation finally promoted to the shipped kit (v1 drums/bass opened −65/−52 dB
+  below peak); instruments re-cut from a roformer-instrumental→demucs CASCADE
+  (no rap ghosts: source windows carried vocals at −23 dB).