I followed one mounting plate through a failed check and reviewed retry, then found its successful result again after our FR10 dashboard restarted; the screen counts 53 passes against 54 planned plates, but our disposition list only accepts 47, so I cannot use its almost-finished message.
Do the results retain a plate identity as well as an attempt identity? You need both if the same plate can have several checks. The count should follow the accepted plates, while those checks remain available underneath.
Both are present: 47 accepted plates, three held, two rejected and two not yet checked; the extra six passes are replayed results, not new inspections.
Our operators used to read nearly complete as permission to clear the bench, even with held parts still there; I would take that message away while you fix the count, not just put a warning in the office spreadsheet.
The misleading completion message is removed, and the temporary view shows the reconciled dispositions with the two unchecked plates still visible rather than quietly treating them as failures.
Try restoring from the saved data as well as importing the same file twice. A counter can look fixed until the old stored total comes back. Derive it from the item decisions when the view is rebuilt.
The developer found exactly that second path: normal imports now ignored duplicate result records, but opening an archived session restored the old accumulated total; both views now rebuild from the plate dispositions, with failed and retried attempts retained.
They can open the held-item list; we checked all 54 plate identities against the disposition sheet, then repeated the imports, restart and archive restore on the installed reporting setup, and it stayed at 47 accepted with every genuine attempt still accessible.
That answers the reporting fault. Keep those replay and restore examples in the tests; the six extra pass rows were useful evidence once you could connect them to the real plates.