Rungs 2-4 embodied as hospital-white placeholder bodies (warden / gunship /
auditor), the gunship's ghost-mirrored sector + parity tether, the auditor's
0:00 stasis bleach, compliance placards, and the v9 onEvents migration.
The inversion is measured, not asserted: fed a worst-case integer-stepped
position that the sim moves on 6 of 120 frames, the warden renders 120 distinct
positions with zero stalled frames. The auditor renders 1 position and 119
stalled frames, deliberately — it moves only between frames (codex 5).
- correction.ts: tick-lerp + damped follow for wardens/gunships, instant
relocation + blink for auditors, instanced ghost mirror (cap 512), tether,
bleach + 00:00:00:00 timecode, and the frozen-entity set
- entities.ts: compliance placard + total idle/mixer/haze freeze (idle clock
delta measured at exactly 0.000000 over 30 sealed frames)
- beltitems.ts: cargo pinned MID-TILE inside a lock; the freeze path is
cheaper than the normal path (0.392 vs 0.465 ms at 2,000 items)
- index.ts: onEvents buffered and drained at the top of render(); repaired,
the firewall can, and relicFound all retired as snapshot heuristics
- notice.ts: the fax-side notice halo, with a one-frame retry because main.ts
runs the renderer BEFORE the UI and the panel may not be laid out yet
- correction.test.ts: 9 tests, mutation-checked. Pins the cross-lane rect
convention against LANE-SIM's own frozenEnt formula over 4,896 cases.
Perf at a deliberate worst case (500 entities, 2,000 items, the entire base
both mirrored AND frozen): whole frame 2.30 ms median against a 16.6 ms
budget; this round's net cost ~0.2 ms. Prod bundle clean of all scaffold.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
462db9947c