Feedback to the distribution-centre builder

This is a public redacted copy of the first repair request. The eight finding descriptions and requested repairs below are unchanged. Local file locations, an internal approval phrase and execution-location instructions have been removed. The original message is retained with the experiment.

The request asked the original producer to preserve its first delivery and return a complete repaired package. See the finding and repair ledger and before/after scripted results for the selected evidence included here. “Independent reviews” in the original message means separate review agents; the trial was not blinded.

All 24 scripted checks mapped from the brief passed. The general baseline returned REJECT with 51 of 52 total rows passing. Independent reviews are separate from that scripted verdict. Address the following items and include a response to every ID (fixed, disputed with evidence, or unresolved):

F0-01 — Selected delivery-policy compatibility. The baseline NormalsExistChecker reports all 121,602 polygon meshes lack authored normals with subdivisionScheme none. Authored normals were not explicitly required in B1: this is a compatibility request for the selected policy, not a claim that your USD is invalid or that you ignored the brief. Author valid normals matching the intended geometry and orientation. Do not merely switch subdivision to silence the rule. Update generation sources and your own QA so the delivered scene reproduces the fix.

F0-02 — Diverter route-state concern. DIV01 remains +35 degrees throughout 10–19 s while P02 passes its JCT03 junction straight toward R01: leading-edge arrival 14.754547 s, centroid 15.272727 s, trailing-edge clear 15.790907 s. Make each traversed junction’s visible state agree with that parcel’s intended route and hold until clear, allowing the previous parcel to clear. The paddle bottom is 158 mm above parcel top; your non-contact kinematic choice is acknowledged. Make its guide-versus-indicator role visually and textually clear without introducing a contact-dynamics requirement. Preserve the brief’s parcel events. If you dispute this observation, provide source-backed reasoning.

F0-03 — Clear-height interpretation. The structural roof starts at 14.000 m but sampled luminaires start at 12.060 m and trays at 12.3775 m. B1 says 14 m clear internal height. Resolve this explicitly, preferably by providing 14 m below overhead services with a correspondingly supported envelope. If you retain a different interpretation, state the exact actual clearance and explain why. This is an interpretation concern from separate review, not a scripted failure or permission to silently weaken B1.

F0-04 — Reframe the saved diverter overview: its overhead tray obscures the mechanisms. The existing close-up and top views partly compensate. Use a suitable view below the obstruction, or a clearly documented inspection visibility state.

F0-05 — Add plausible mounting for the maintenance sign floating above G01, or move it onto supported geometry.

F0-06 — Complete the status tower mast/collars. This was already disclosed by you, not newly discovered by the harness. Keep the exact stepped one-lamp state schedule intact.

F0-07 — Improve sequence readability with a closer synchronized view or explanation overlay that makes parcel IDs, scan events and lamp states understandable. Retain the wide facility sequence; do not claim a measured timing defect where the scripted timing checks passed.

F0-08 — The required actual reopened hierarchy/timeline screenshots and observed interactive navigation/playback performance remain missing as you disclosed. Supply genuine captures and observations if your available environment supports them. If unavailable, explicitly retain this limitation; do not fabricate GUI evidence or label CPU transform evaluation as interactive performance.

Deliver a full portable V1 scene, updated sources/inventories/scene-map, your validation, exact per-finding response, revised rendered stills/video sufficient to review changes, renderer/revision provenance, and a new manifest. Preserve already-passing requirements, including prim count, dimensions, supplied image pixels/UVs/materials, metadata and motion. Record actual effort/timings when available. No need to read or modify evaluator implementation; the reports and reviewer records above supply the observations.