Why does a reviewed retry make two accepted brackets?

Bracket identities are persistent, but the dashboard sums successful attempts; a reviewer can also put an item back on hold after a pass, which that sum ignores.

Then even counting distinct brackets with any pass would be wrong. The current disposition needs to drive the accepted count. Sorry, I should have asked about hold and re-review before assuming the only issue was duplicate retries.

Show attempts and accepted items as different numbers, with labels people can tell apart

And don't make a retry erase the failure. The item can be accepted after review while its history still records why it needed another inspection. Those are different facts, both useful to the people running the process.

Developer has proposed a current-disposition view per bracket, with the attempt history unchanged; I have added accepted-to-hold and repeated-success cases to the review.

Include a restart while a disposition update is being saved. The screen must not claim a decision was committed when only the last inspection event reached storage. Let uncertain records stay visibly unresolved until reconciled.

Who settles an uncertain record when the reviewer has gone home?

Name that owner in the shift handover, with access to the original attempt and review records; don't let the next operator invent acceptance to clear a queue.

Has the revised count been compared with your reconciled batch yet? I'd be interested in whether holds account for some of the discrepancy as well as the extra successful attempts.

The offline comparison now matches the reconciled item dispositions, including holds; the original attempt totals remain available, and restart handling still needs its acceptance check before deployment.