pending has become a dangerous little word in our startup helper

Before the request call. So some pending jobs never left us, and others could have finished. We can't separate them from that field.

Treat those existing ambiguous entries as requiring reconciliation, then distinguish never submitted work from an attempt whose outcome is unknown in the revised records.

Who will investigate an unknown entry when the developer isn't around?

Maintenance with the process owner. I'm the lab contact, not the person who can clear an unknown inspection from a hunch. We're writing that route into the handover.

Does each submitted attempt have an identity separate from the plate identifier?

Only the plate label at the moment. Developer is adding an attempt identity and keeping the plate link. We'll also check what identifying evidence the receiving side retains.

Good, but don't let the new field make the old records look certain. They still need the investigation you described.

Let people open the history without clearing the hold. A view button that quietly tidies up state is a surprisingly effective way to lose the awkward evidence.

Viewing and reconciliation are separate actions in the draft. Old entries stay held. Nobody has retroactively assigned them a successful outcome.

Does refreshing the page trigger the startup helper again? Ours did something similar once.

It did indirectly. The refresh route called the same queue-resume function. Developer has removed that connection as well. Glad you asked before we only fixed application launch.

Will the tests interrupt execution around both the attempt record being saved and the request being sent? There can be uncertainty even when the call never returns.

And after the result arrives but before it is saved. That's another place the screen can know more than the file.

Those cases now stay held in the offline tests. A refresh or restart doesn't resubmit them. We also tested the attempt marked but never sent; it stays uncertain until reconciled, which is annoying but deliberate.

What if the plate has been removed before maintenance gets there? Your electronic entry may survive longer than the physical arrangement it describes.

Then the procedure must use the retained identity and available evidence, not assume the current fixture contains that plate. If they can't establish the outcome, it remains unknown.

Can only the authorised reconciliation role release it from that hold?

That's the intended permission. Maintenance and the process owner are reviewing it with the developer. Offline restart behaviour is improved, but the complete recovery route isn't signed off yet. Thanks for catching the refresh path, Mei.