docs(board): archive the OPAL rig-hardening tasks resolved 2026-07-28/29

#46 #90 #94 #97 #100 #84 #35 off the active board, each with the non-obvious part
kept: the 32 role inversions phase 2 shipped and how they were repaired, why the FX
bus rule is PV011 and not PV010, and the two independent highlight bugs (a class name
that never matched, and an outline fix written in the one stylesheet copy that could
not unset the other's border). Narrative in log 019.

armada/tasks/board-archive.md

@@ -2,7 +2,7 @@
 
 Snapshot of resolved Claude-Code board tasks, archived before /compact so the active
 board stays lean. Narrative achievement logs live in the numbered `NNN-*.md` files; this
-is just the tracker history (id · subject). Last archived: 2026-06-07.
+is just the tracker history (id · subject). Last archived: 2026-07-29.
 
 ## Foundation & distribution
 - #1 Scaffold L'Armada toolbox in Tidal repo
@@ -58,3 +58,24 @@ is just the tracker history (id · subject). Last archived: 2026-06-07.
 - #88 Landing index + track-style tag cloud
 - #64 EPIC: 'Information is Beautiful' dataviz of the corpus — shipped (corpus.html
   narrative + Unwrapped explorable w/ vibe-search, density field, zoom + landing)
+
+## OPAL 2026 rig hardening — the surface, and the transition (log 017, 018, 019)
+- #46 SURFACE REMAP phase 1: faders + knobs column-aligned
+- #90 REMAP verification gate PASSED for phase 1
+- #94 REMAP phase 2: the button rows column-align (166 refs) — then the 32 role
+  inversions it shipped, repaired the same day by tools/button_roles.py + the
+  role-first migrator (5089eec, 2c178c9, 6ed637d).  Left open on purpose: 12 sites
+  where one orbit drives two gestures and only one momentary cell exists (needs #54).
+- #97 ONE authored grid: five hardcoded copies of the LCXL convention become derived
+  or generated (2778f1c)
+- #100 pvlint PV011 — one FX bus SLOT shared by two orbits ("like shared busses they
+  screw things").  Landed as PV011 because PV010 took the button role/row rule PLN
+  asked for in the same breath.  Found crushbus 41 shared by take_5's bass and choir.
+- #84 the editor now TELLS the rig which track is loaded — the HUD writes
+  ~/.cache/parvagues/current-track, which lcxl-leds --watch already polled but nothing
+  ever wrote during a set (2576e77, pulsar-parvagues-hud)
+- #35 eval-highlight: per-stream colours never matched (runtime emitted
+  `event-highlight-1`, CSS said `.event-highlight-d1`) AND the wiggle was still live,
+  because the outline fix sat in the ~/.pulsar/styles.less copy while `border: 1px`
+  survived in the package sheet underneath — a cascade winner cannot unset a property
+  it never mentions (2c0524e, pulsar-tidalcycles)