Saved history contains it with the job identity. The connection handler creates an empty list of seen completions on every reconnect.

Reconcile against the durable history rather than the connection's temporary list. Reconnection is a transport event; it need not create a new inspection or erase knowledge of a completed one.

And does anyone have to press continue? That word can mean almost anything after an interruption.

It currently shows connected and the old total immediately. No operator button. Developer is adding a reconciling indication before the current total is presented.

What happens if the retained completion has an identity absent from the saved history? It could be work completed while the monitor was away.

That case needs a defined reconciliation rule. Absence from the local completed set is not enough to treat an arbitrary retained state as a new authorised result. Compare it with the persisted pending job too.

Theo, would you show that unknown case separately from the duplicate one? For a covering operator, one is already accounted for and the other needs somebody to decide what happened.

Yes. A known duplicate should settle without requiring the operator to untangle it. An unmatched retained result should remain visible as unresolved, with its identity available to the responsible person.

Revised tests distinguish already-recorded completion, completion matching a pending job, and unmatched retained state. No clearing or resubmission used to make the examples pass.

Who takes the unmatched case when you're absent, Sara? A nice unresolved label can still leave the next person stranded.

Shift lead uses the agreed reconciliation route with controls when needed. The walkthrough found that contact missing from the page; it is there now, beside the unresolved job identity.

Have repeated reconnects stopped increasing the count yet?

Yes. Repeated reconnects and full application restart preserve the correct total from saved history. Pending-job completion records once; unmatched state stays unresolved. The tested monitor correction is installed and this duplicate-count fault is closed.

Thanks for including what the covering operator sees. That's usually the detail I have to ask for after the software fix is declared finished.