Playtester report: (1) the CONTROL MODE stack showed boxes around the inactive
MID/ADV entries, (2) a phantom block between the HEADING dial and the ARMOR
rosette. Root-caused [T1] by a 4-agent investigation + art decode:
The #48 translation-table cycle-fill (6bb03ae, 2026-07-25) mapped
out-of-range art indices in-plane as (index mod 2^bits) -- so art index 254
landed on plane slot 62 = the LIVE colorMapperMultiArmor right-armor damage
slot. And idx-254 art EXISTS: every SMODE.PCC frame fills the INACTIVE
mode-box interiors with 254, and BTSEC1.PCX carries a stray 52x13 idx-254
bar at port (199-250,101-113) between the heading dial and the rosette (a
scratch duplicate of the rosette quadrant bars). Both lit up in the current
right-armor color (adpal ramp green/orange/red) on EVERY render path. On
the shipped machine those regions rendered BLACK (the garbage entry's low
plane bits were 0), which is why the 2026-07-19 smode audit -- run before
the cycle-fill landed -- verified CORRECT.
FIX 1 (BuildSecondaryTranslation): map [2^bits..255] to translationTable[0]
(plane BACKGROUND) -- the authentic on-screen result, same no-leak
guarantee. Verified on both paths: MID/ADV back to authored borders+text
(idx 5/9), no interior fills; the phantom bar gone; BAS badge and the live
armor rosette (in-range slots 60-63) untouched.
FIX 2 (glass dirty-skip palette blindness, latent in c9e25e5): the
ColorMapper family animates by CLUT writes with zero pixel churn, which the
pixel-checksum token cannot see -- the glass radar held stale armor tints
between pixel repaints. New SVGA16::paletteGeneration, bumped by the
palette writers (BuildSecondaryColor only on a REAL entry change -- the
adpal/adpal2 flash alternates every Execute writing identical RGB at zero
damage, so the skip survives; full rebuilds bump unconditionally), folded
into GlassWindowToken for palette-expanding windows.
KB: gauges-hud #48 REFINED addendum; glass-cockpit repaint-pump section
(records f3d27f5 + c9e25e5 + this); GAUGE_COMPOSITE row 33 corrected
(ControlsMapper/ControlMode, not DisplayMode) + re-verification note.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>