Python only wrote the reused label. PLC trace includes a request counter. I'm looking for somewhere that counter reached the application log.

Check the raw response capture too, if you retained one.

Yes, keep each source in order before trying to align them.

Found a raw capture from before the restart. It contains the PLC counter beside the first label. Nothing equivalent after restart. At least the first request has an identity now.

Distinguish intended work from transmitted work. An application retry entry does not establish that a second request reached the PLC. Preserve that uncertainty in the reconstruction until a corresponding transition or message can be identified.

Does the PLC counter advance after the restart?

No advance in the captured interval. And the application line says 'retry pending', not 'retry sent'. I shortened that too far in my opening.

Then a second executed attempt is not established by those records.

That removes one mystery cheaply. Still need to tie the acknowledgement to the first request, though.

The PLC trace has acknowledgement changing after that request counter appears. The application saw it later, following reconnect. I'll describe those as separate observations instead of two acknowledgements.

What will the incoming operator see on the job screen? The log explanation is improving, but they need to know whether that coupon's result is usable without becoming the fourth person interpreting three clocks.

An acknowledgement establishes only what the interface defines it to acknowledge. If it means request acceptance, it does not establish inspection completion or a passing result. That distinction should determine the displayed status.

What does your interface document say this acknowledgement means?

Request accepted. No result payload survived in our capture. The screen has been marked awaiting reconciliation by the controls lead; the coupon isn't being counted as inspected from this evidence.

Worth adding the request counter and application startup identity to future logs. Doesn't mend this record, but avoids the same guessing exercise.

Added both to the logging change request. Handover now says one confirmed accepted request, completion unknown, and no confirmed second request. That's less dramatic than my original version, fortunately.

That's a handover someone can use. Who covers the reconciliation if the controls lead is away when the next shift starts? The named contact needs a backup, especially for a state that blocks normal work.