docs(tasks): archive #66 — six wrong instruments and one solid rig

The full #66 entry for the documentary trail. The headline is not "all 10 tracks play"
(they do, 91 orbits, zero silent) but that the instruments were wrong more often than the
rig — six times, and the arithmetic was never the bug in any of them. Three answered a
question about SHAPE with summary statistics, one divided by zero and called it an effect,
one compared two points on the same hill, and one measured a system that included the
measurement.

The last is the best story: an xrun counter climbing +816 in 25 seconds, days before a
gig, matching PLN's own report of crackles — and caused entirely by probe-chain's own
PipeWire capture nodes. Remove them, repeat the identical musical load, get zero.

Both REAL faults that day were compile errors wearing an audio costume.

armada/tasks/completed-archive.md

@@ -464,3 +464,61 @@ batch of measurements all come back empty, suspect the harness before the rig.*
 **Deps.** Closed #26 and #63. Unmasked #61 (which then measured *clean* for the OPAL set —
 all 41 seeded CCs cover it). Spawned #65 (an undefined `crackles` silencing nine orbits,
 which is a compile error, not audio), #66, #67. Feeds #44 (readiness gate) and #24.
+
+---
+
+## #66 — Prove all 10 OPAL tracks playable + every control audibly effective
+
+**Description.** PLN left the laptop running at J-7 and authorised autonomous work:
+*"ensure all tracks from gig are playable. you can send fake midi events for each, thus
+test all controls have impact on sound ;) even the crushes should change noticeably what
+you see/'hear'"*. No ears in the room — every claim instrumental, every failure reported.
+
+**Done.** `check-tracks.sh` end-to-end: **10/10 tracks, 91 orbits, zero silent, zero eaten
+by an Ardour fader.** Two were broken at the start of the day — `desire.tidal` at **0/9**
+(`off 0.125 (|+ note 12 . (|* gain 0.6))`: `.` binds tighter than the section, so it
+composed a function into a ValueMap; one pair of parens → 9/9, `cfa56b6`) and
+`vague_de_crime` (PLN's own `crackles` fix, saved). Every control measured effective with
+the lens that suits it: `gF1` centroid 112→3681 Hz, `gF3` 384→2765 Hz (+529%), `gMute2`
+−23 dB→silence, `crushbus` 125→189 Hz (+37%), `ply ^89` 8.1→4.7 onsets/s, `gMask` −33%
+onsets, `gMute1`+`gM1` silencing four orbits at once. `ply ^91` on d7 reported
+**INCONCLUSIVE** and left that way. Shipped `tools/control-lens.py`,
+`tools/check-tracks.sh`, `tools/tempo-lens.py`; commits `cfa56b6` `708c2ae` `7b2e5b4`
+`16429b0` `36466a4` `4836d74` `104fb03` `156444d`.
+
+**Learnings — the instruments were wrong more often than the rig, six times, and the
+arithmetic was never the bug.**
+1. *rms is blind to a filter or a crusher* — they move timbre at constant loudness.
+   `gMask` read "NO CHANGE, 0.5 dB": right sum, wrong question; on the density lens, −42%.
+2. *`ply` reported "+inf %, EFFECTIVE" off a division by zero* — the envelope onset
+   detector read 0.0/s on a sustained bass. The tool invented an effect, the one thing it
+   promises never to do. Fixed with spectral flux + an INCONCLUSIVE floor.
+3. *A decay crosses its own midpoint exactly once.* A healthy `mask "<t f!7>"` riser was
+   called "DECAYING, almost certainly an xfade tail" twice. Two quartile averages cannot
+   answer a monotonicity question; counting crossings can.
+4. *One low FINAL bin is the window closing, not a fade* — `perfect`'s d4, steady for
+   twelve bins, condemned on the thirteenth.
+5. *A uniform +1.8% tempo bias* came from a threshold rule sliding down the left flank of
+   the RIGHT peak — comparing points on one hill as if they were different hills. (I first
+   blamed `np.correlate`'s unnormalised overlap and fixed that too; the bias didn't budge.
+   **Fixing a real-but-irrelevant flaw is not evidence you found the cause.**)
+6. *The observer was the effect.* Ardour's xrun counter climbed +816 in 25 s under the ply
+   stress test — which reads exactly like "heavy patterns cause dropouts on gig week" and
+   matched PLN's own crackle report. Remove the ~18 PipeWire capture nodes probe-chain
+   creates, repeat the identical musical load: **zero xruns**, scsynth ERR = 1 all session.
+
+*And the discipline that makes the rest survivable:* **A-B-A**. These patterns vary per
+cycle by themselves (`d2` of `vague_de_crime` plays a different bar every cycle), so
+re-measure the baseline at the END and treat any swing under 2× the self-drift as
+inconclusive. Plus: take the baseline **twice** — a single closing repeat agreed with the
+opening one by luck and reported drift 0.0% when the true within-condition spread was 33%.
+
+*Also:* `you_my_sunshine`'s alarming "144 vs 166" tempo mismatch was **a commented-out
+`setcps` line** — a parser miss masquerading as a data conflict, again. And the RUNBOOK's
+pre-set checklist had been telling PLN to put the DJ filters DOWN for bypass; **centre is
+bypass**, down is a hard low-pass.
+
+**Deps.** Closed #65. Fed #62 (the live mixer is clean; the SAVED file still isn't),
+#56 (event density exonerated as a crackle cause), #12 (one of three tempo questions
+dissolved), #33. Spawned #68 (orphan orbits). Ran alongside #67 (LED painter) in a
+parallel agent.