Attempt identity survives in the PLC record and application's saved result. They agree for this occurrence. Application increments on its first high read without checking whether that attempt was already recorded.

Then the recorded attempt can be reconciled without creating another one. Keep an unmatched-identity case in the repair tests as well; your complete evidence here should not become a rule that every retained high is a known repeat.

Developer's offline repair now checks the durable attempt record before changing bookkeeping. Known repeat leaves the total alone; unknown identity is held for review. No automatic dispatch in either case.

Did reopening reconstruct the same result and preserve the unmatched item, or did the test stay in one running session? That matters for the shift that restarts the monitor after an outage.

Reopen preserves both in the test. The identified old completion is reconciled, but installed-interface checks and the review handover for unmatched cases are still outstanding. Repair not released live.

Keep the retained-state capture with that test result. It explains why a transition-only approach failed and gives the installed check a specific recorded case to answer, rather than a general claim that reconnect works.