tide-table: gig-eve rehearsal notes, six tracks and four ear-signals

PLN played the arc to take_5_drops and called it: 'not bad i am confident'.
That is the only confidence signal this set gets before doors, so it is
logged with what it covers (six of seventeen) and what it does not.

Four reactions paired with their mechanical causes, which is the point of this
corpus: piment_bresilien's silent d9 (a bank off SuperDirt's load path, plus a
parser that read 90s_synatm as s_synatm and 808bd as bd and so warmed the
WRONG banks for five of seventeen tracks); crackles he heard in one listen
that four logged sessions had not surfaced, attributed by xrun_by to Ardour's
GUI thread on a realtime policy; salut_nu having no backdrop because
defaultScene ships empty, which reads as the same taste as 'the midi mon is
beautiful' -- the visual is part of the instrument; and the DJF third row
landing despite his own bezel finding, which refines that finding rather than
contradicting it.

armada/tide-table/performance_notes.md

@@ -664,3 +664,82 @@ outcome — "works great" — only arrived after the *wiring* was right, which i
 the standing lesson: a taste complaint (or its absence) tracks a mechanical
 cause. See also: the double-CC path (`MIDIIn.connectAll` with an unfiltered
 `src`) that would have made toggle buttons silently no-op.
+
+---
+
+## 2026-09-24, 00:00–01:15 — gig-eve rehearsal on the XPS24 (not a take)
+
+The night before "AI ENGINEER". Not a recording: PLN played the arc on the rig
+while the rig was being repaired underneath him, so every reaction below is
+paired with a mechanical cause that was found the same hour.
+
+**He got through the first third and he trusts it.** *"i did rehearse up to take
+five drops not bad i am confident"* — that is six tracks of the seventeen
+(`salut_nu`, `liquid_nite`, `love_first`, `piment_bresilien`,
+`mafia_sans_serif`, up to `take_5_drops`), played for the first time in sequence,
+and the verdict on the ARC rather than on any one track. Record it as the only
+confidence signal the set will ever get before doors, because the 2h rehearsal
+never happened and the 5–6pm window is one hour for a two-hour set. Tracks 7–17
+are un-rehearsed by his own account.
+
+**`piment_bresilien` d9 was silent, and it was not the score.** *"i tried
+piment_Bresilien dont hear d9 i see its events HL tho and the fader is up but no
+sound in ardour either. is 90s_synatm:10 samples missing?"* — a textbook read of
+the symptom, and the right instinct. The bank was never missing: 36 files sat in
+`Samples/90sArchive/output/90s_synatm`, symlinked from `samples-links/`, which is
+**not** one of the three roots SuperDirt loads (`Dirt-Samples`, `Samples/extra`,
+the drum machines). `note` produced events, so the HUD lit; no buffer existed, so
+nothing reached Ardour; the fader was never in the story. Fixed with a link into
+`extra/`, and confirmed by ear after the SC restart: *"thanks i hear piment d9
+now :))"*.
+
+  **Sampling lesson, and it is the durable one.** The preload had been warming
+  the WRONG banks for five of the seventeen tracks. `extract_names` required a
+  letter first, so `"90s_synatm:10"` read as `s_synatm` and `"808bd"` read as
+  `bd` — and `bd`, `hc`, `sd` are all real Dirt-Samples folders, so the
+  truncation *resolved* instead of failing. Any bank whose name starts with a
+  digit was invisible to every tool that asks "is this sample warm". Measured
+  gain across the Thursday set: `808bd` x3, `808cy`, `808hc`, `808sd`,
+  `90s_synatm`. When PLN says a sound is missing, check what the PRELOAD thinks
+  it is called before checking the disk.
+
+**Crackles, heard and then attributed.** *"also just heard crackles, are they
+skips?"* — yes, and gig-log answered which process: `xrun_by.ardour` went **0 ->
+46** in the six minutes after Ardour opened, while SuperCollider added 9. Cause
+was Ardour's GUI thread on `SCHED_FIFO/80` (put there by `parvagues-protect`)
+burning 40–85 % of a core. Chronic, not new: 5172 / 4852 / 1114 / 987 ardour
+xruns in the Sep 10 / 18 / 20 / 22 sessions, four sessions of evidence nobody had
+ever read by name. **After demoting those threads: 0 xruns in 14 minutes**, at
+the same 65 % clock cap — so the cap may not be the binding constraint at all.
+His ear caught in one listen what four logged sessions had not surfaced.
+
+**`salut_nu` has no backdrop, and that is a config gap not a bug.** *"starting
+the set moving from rose_rouge.tidal practice to salut_nu, no more scene? why no
+default scene?"* — `rose_rouge` carries `-- url: rose_bloom.gif`; `salut_nu`
+carries no `-- scene:` directive, and essentially none of the 716 `.tidal` files
+do. The HUD has a `defaultScene` setting that ships empty. Taste signal worth
+keeping: he expects a backdrop to be the **default state**, not an opt-in per
+track — the visual is part of the instrument, the same read as "the midi mon is
+beautiful" above.
+
+**Surface taste: the DJF third row lands.** *"row 3 wunderbar"* — the OLED's
+third field now carries travel from the bypass detent as a signed percent
+(`-100% … 0% … +100%`), each side normalised against its own throw. He had asked
+for it on row 3 specifically, against the 2026-08-29 note that row 3 hides behind
+the bezel from his chair; with the board in hand he kept it. So the bezel finding
+holds for *the row that matters* (the cutoff, row 2) but a glanceable number is
+welcome in the corner. **Two-finger one-hand DJF reset approved** on the same
+reasoning: *"this allows fast reset DJF with one hand 2 fingers"*, with arm-light-
+and-OLED-before-drop as a condition, not a nicety.
+
+**On trusting instruments.** *"i not sure i can trust"* (about the CPU perf
+numbers) and *"dont rely on my paste nor memories"*. Both are instrument
+complaints, not mood: `gig-up` had been printing "thermal-mode performance
+holding" from a config file while the gate, twelve lines below in the same log,
+failed on the live knob. He is right that a number he cannot verify is worse than
+no number. The launcher now reads the knobs and names their owner.
+
+**He stopped while ahead.** *"im tired imma fold"* at 01:10, six tracks in, after
+a thirteen-minute silence caused by a diagnostic SuperDirt restart. Worth
+recording as pacing data: the gig-eve session that produced the most repairs also
+produced the least playing, and he called it himself rather than grinding.