File integrity is only one recovery gate
The snapshot, retained exceptions and narrow dependency restore had measurable outcomes; application recovery remained open.
Published 2026-11-01 · Dated evidence, not a live status feed
What was preserved and compared
The local payload reached 87,403 files after recovery of 379 initial long-path failures. Twenty sample restored files matched their source hashes. Ten project delivery archives were copied and checked; six asset parts restored 262 files with matching manifest identities and hashes. These were file-preservation results, not an application restore demonstration.
Keep the ledgers distinct
The corrected main total was 87,880 records: 87,857 passing and 23 failed. An earlier short subtotal omitted 92 records; the clarification corrected the arithmetic without rerunning the whole audit.
The 23 failures concerned metadata-stat stability, while content bytes and final hashes matched. The original before-stat values were not retained, so the historical discrepancy could not be reconstructed. Later diagnostics passed; they did not erase the failed rows or prove that 23 files had been corrupt.
| Evidence set | Recorded result |
|---|---|
| Main audit | 87,880 records; 87,857 PASS / 23 FAIL |
| Later metadata diagnostic | 63 PASS; original exceptions retained |
| Separate narrow follow-up | 6,330 PASS across 2,110 files |
The narrow restore closed a data gap
A static scene/dependency review found that the small restore sample lacked the DDS edit’s 2,034 image dependencies: 1,998 preserved frames plus 36 camera patches. The narrowed follow-up copied those images and 76 ordinary static artifacts. Its separate preflight, size/hash and format-integrity checks all passed.
The copy ran at 00:22–00:23 UTC on October 25, equivalent to 20:22–20:23 Eastern daylight time on October 24. The public event date follows the latter; UTC-named folders do not shift it to the next local day.
What a passing copy did not prove
Scene paths still pointed to the original media locations. Isolated opening, relinking, playback, cache use and a representative render were still required. A byte comparison or image decode does not prove an application will recover its editing state.
Conversation-history preservation, unsaved application state and a tested complete recovery path were not established. Reset remained NO-GO in the retained record. This publication performs no reset, restoration or new device test.