docs(tasks): archive #62/#113 — the denominator that moved a task, and the 108 idle slots

armada/tasks/completed-archive.md

@@ -1266,3 +1266,80 @@ invisible forever after.
 
 **Deps.** Blocked nothing; it was the difference between a suite at 213/1-fail and a
 clean baseline to migrate against, which is why it went first.
+
+## #62 — Ardour faders: the SAVED session and the LIVE mixer disagree
+
+**Description.** `check-mix.py` kept reporting Tidal tracks at -inf dB in the saved
+Ardour session while the live mixer sounded fine. The class of bug that cost PLN a whole
+evening on 2026-07-27, when "d7 fades to nothing" and "d9-d12 make no sound" were chased
+through Tidal, SuperDirt and PipeWire before anyone looked at the fader.
+
+**Done.** Closed 2026-07-31 by PLN raising Tidal 10 and saving. `check-mix.py` reports
+no failing orbits.
+
+**Learnings.** Three things, and the middle one is the reusable one.
+
+1. **Reframing moved it.** On 2026-07-30 it read as "3 faders are down" and sat mid-board.
+   Cross-referencing the silent tracks against the orbits each setlist track actually
+   declares (pvlint's parser, not a grep) turned it into **"9 of your 13 tracks would
+   lose at least one orbit on a relaunch"** — and it became the top item. The measurement
+   didn't change; the denominator did. Worth doing to any infrastructure finding before
+   deciding it's minor.
+2. **A save splits the diagnosis in two, and you only see it if you re-measure AFTER.**
+   PLN's Ctrl+S fixed Tidal 06 and 12 but *not* 10. A fader still at -inf after a save is
+   not an unpersisted fix — its live value is genuinely down. So there are two remedies
+   hiding under one symptom: "Ctrl+S" for a fader that disagrees with the desk, "raise
+   this fader" for one that survives a save. Re-running the check after the fix is what
+   separates them; without it, the second class hides behind the first indefinitely.
+3. **It took three sessions with a different set of tracks each time** (07-28: 05/06/08/10;
+   07-30: 06/10/12; 07-31: 10). Root cause is structural — the D-row faders are
+   MIDI-learned to CC 77-84, so the physical desk and the session file drift
+   independently, in both directions. That is why it must become a hard fail in the #44
+   pre-gig gate rather than a thing to remember.
+
+Never fixed from outside: writing faders desyncs them from the physical desk, CC77 is
+global gain where down is total silence, and levels are musical intent. The ask was
+always one keystroke, and framing it that way is what got it done.
+
+**Deps.** Feeds #44 (must be a hard fail). Was the top blocker on trusting any launch.
+
+## #113 — Set coherence review: unmapped effect/button slots across the set
+
+**Description.** PLN, walking the OPAL set by eye: *"i saw some unmapped effects/button,
+a 'set coherence mapping review' and quick wins suggestions (bass octersub, lead slice,
+etc etc) could be good ways for you to fill-in suggest"*.
+
+**Done.** 2026-07-31, commit a32c6d2. `tools/set-coherence.py` +
+`docs/2026-07-31-set-coherence.md`. **108 idle slots across 117 declared orbits.**
+
+**Learnings.**
+
+- **Absence has no line number.** You cannot grep for a control that isn't there. The
+  only way to see an unused affordance on a fixed grid is to compute `OWNS − USES` per
+  orbit — `lcxl_grid.orbit_home(n)` for the cells, pvlint's parser for the block, `"^NN"`
+  refs for usage. Generalises to any "what's missing" question over a known schema.
+- **Report both directions of a coherence question**, because fixing one creates the
+  other: IDLE (owns a slot it never uses) was uncovered until now; STARVED (drives more
+  than it owns) was already PV008 + the FIXME(#54)/(#94) markers.
+- **Three systematic findings beat a flat list of 108.** (a) The drums' own effect knobs
+  are the deadest cells on the board — d2.fx idle 12/13, d1.fx 9/13, d3.fx 8/13, and
+  since d1-d3 own exactly ONE per-orbit knob each this is their only knob, unused; the
+  kick has no per-orbit effect knob bound anywhere in the set. (b) Row A (d9-d12, A5-A8)
+  ~90% idle, which is #98's pooling proposal with its evidence attached. (c) d6 has gates
+  but no knobs in 5/5 tracks — fallout from the d7→d6 consolidation, where ^58/^90 came
+  along and ^34/^54 did not. **An orbit renumbering carries the bindings it names and
+  silently drops the ones it doesn't.**
+- **The finding that inverted the ask.** Measured what PLN actually binds to a per-orbit
+  effect knob — perc `legato`(4); bass `lpf`(3), `gain`, `bandf`; lead `width`(2),
+  `modIndex`, `filterRange`, `cutoff`, `resonance`; kick nothing — and his two *named
+  wants sit outside his own idiom*: he binds lpf to bass knobs and never octersub, and
+  slice lives on his buttons not his knobs. So "bass octersub, lead slice" are additions
+  he is reaching for, not drift being corrected. Telling him that is more useful than
+  pretending the corpus implied them, and it is a step worth adding to
+  self-reference work generally: check whether the user's own request matches their own
+  history before treating it as a restoration.
+
+Nothing applied — every suggestion changes how a track sounds and he rehearses on these
+files. Also confirmed the #94 remap held: borrowed CCs are 0 or 1 across every track.
+
+**Deps.** Confirms #98 with data. The IDLE counterpart to #54's STARVED problem.