writeRound (commit.ts:977) writes the round's EDL, then the auto-cut fold rebuilds and writes over the same file. After a commit, the round directory holds only the post-fold EDL.
Consequence: the deterministic pass's own effect is unauditable after the fact. Verifying #66 turned up a 5.615s discrepancy (547.900 - 136.486 = 411.414 expected, 417.029 measured) and it could not be explained, because reconstructing the pre-fold EDL to diff against was no longer possible. The likely cause is the per-segment edge fade re-added on rebuild, but that stays a hypothesis.
auto-cut.json records the cuts it emitted, which is not the same thing as what the EDL looked like before they were folded in.
Keeping the pre-fold EDL under a distinct name would make rounds --diff able to show what the deterministic pass actually removed, and would let a duration mismatch be traced instead of guessed at.
writeRound(commit.ts:977) writes the round's EDL, then the auto-cut fold rebuilds and writes over the same file. After a commit, the round directory holds only the post-fold EDL.Consequence: the deterministic pass's own effect is unauditable after the fact. Verifying #66 turned up a 5.615s discrepancy (547.900 - 136.486 = 411.414 expected, 417.029 measured) and it could not be explained, because reconstructing the pre-fold EDL to diff against was no longer possible. The likely cause is the per-segment edge fade re-added on rebuild, but that stays a hypothesis.
auto-cut.jsonrecords the cuts it emitted, which is not the same thing as what the EDL looked like before they were folded in.Keeping the pre-fold EDL under a distinct name would make
rounds --diffable to show what the deterministic pass actually removed, and would let a duration mismatch be traced instead of guessed at.