# Board commitment and connection rebuild: revision r03

**Changes required. This is a packaged candidate, not a finished fidelity study.**
One source treatment has two original-content Remotion renders. The Resolve
instructions describe manual reconstruction; native Resolve was not executed.

## What was checked

The existing independent source-first review inspected ordered source stills
at native indices 2613–2672. The present implementation follows its joint
pool/board/readout commit followed by connection rebuilding. That earlier
review and its findings remain unchanged.

The present builder inspected all 60 selected source indices in six existing
overview sheets, with native-PNG checks at 2641 and 2644. Twelve neighboring
thumbnails were incidental context, not additional treatment coverage. The
builder then inspected all 120 decoded output frames in ten 12-frame sheets.
This is **prior-informed static inspection**, not independent, blind, native
pixel-by-pixel or moving review. No moving-video quality vote is recorded.

Both actual outputs contain 60 frames at exactly 24000/1001 fps, 1920×1080,
H.264/yuv420p, with no audio. Per-frame PTS and durations pass the package's
verifier. Twenty-five tests and TypeScript checking pass. Both 60-row Resolve
tables exactly regenerate from the code used by the renderer. Tests used
already-installed pinned dependencies; a fresh dependency install was not run.

## Repairs visible in these outputs

- Upper committed connection is already partial at local 28, rather than absent.
- Numeric score changes at 26 while the fill bar advances afterward.
- Occupied gold arms differ from the empty/preview lower link; three occupied
  midpoint badges are restored before the commit.
- A local payout label/spark group is separate from later right-panel accents.
- Brown intermediate graph states sit between the observed gold/dark anchors.
  Pale panels, occupancy and committed readout remain unchanged during them.

Top/center labels retain the earlier collision repair. At local 25 the traveler
is absent while the old score/pool/target preview remain; all three commit at 26.
The original and adaptation use the same frame-to-state function.

## Remaining differences

Original schematic artwork replaces the source's illustrated cards, serif
hierarchy, fine material effects and richer local labels. The source's lower
unfilled-arm midpoint badge is still omitted; the last pool card is not centered
as in the source. Slight source board-only vertical shifts are not reproduced.
Connection ratios, bar progression, RGB levels, trajectories and fades are
authored interpolations, not measured source algorithms. The small particles
and glow are simpler, with different counts, shapes and intensity. The separate
region ownership is represented, not exact particle tracking.

The prior five findings are therefore addressed structurally but are not closed
by a fresh independent review. Smoothness, motion blur, transition perception,
source-quality completion and native Resolve parity remain unverified.

## Reuse

Download the complete editable project and change its two `inputs/*.json`
records. Follow `project/README.md`, then `project/docs/RESOLVE-R03.md`. The
older Resolve notes are explicitly historical. No source film, source raster,
logo or demo data is a render dependency. The source excerpt is a separate
attributed research comparison, not an editable asset or narration track.