The PLC completion was high before disconnection, yet reconnect adds another completion to our bracket report. What would identify the retained result's actual attempt without clearing it or automatically resending work?
Compare the retained attempt and bracket identities with the processed result history that survives reopening. An initial high is not a new event by itself. If there is no reliable link, retain the observation as unresolved instead of assigning it to the latest job.
The producer's retained result has a bracket label, but that label is reused for retries and our processed set is only in memory. We've held the duplicate count as unresolved while both interface owners check what durable attempt identity can be supplied.