Yes, labels are stable within the batch; completed attempts currently drive the counter whether or not they add a newly accepted bracket.

Build the total from distinct accepted batch-item records. Keep attempts attached; don't edit history to fit the desired number.

First settle what accepted means after a reviewed retry or a later failure. Otherwise distinct labels alone can still produce the wrong total, just without obvious duplicates.

Who is authorised to change that disposition, and who sees items still awaiting review?

Quality owns disposition; I'm asking for the written rule and keeping pending items visible rather than counting them as finished.

Keep the attempt that led to that disposition identifiable too, so an operator can find the actual accepted check rather than browse every retry

And preserve the other attempts. A link to the accepted one isn't a reason to discard the failures.

For software tests, replay the same events twice and reconstruct after restart. Both routes should reach the same per-item dispositions and total under quality's rule, not merely stop the screen adding twice.

Quality has agreed the rule, including the review hold after a later failure; the revised report derives totals from item disposition and links every attempt beneath its bracket.

Can the supervisor distinguish remaining unaccepted brackets from missing inspection records? Those are different work queues.

Yes, the report separates those states and shows uncertain records explicitly; thank you Anna, our old nearly-done display hid that distinction.

Have you checked the corrected batch against the retained inspection and review records, not just against itself after restart?

Quality reconciled this batch against those records, and duplicate replay plus restart reconstruction preserve the same accepted total; the attempt history remains intact, so we have released the corrected report.

Keep that reconciled example with the tests. It gives the next maintainer an expected item-level answer, not merely a plausible-looking grand total.