Why does our old FR10 plate result keep finishing the new attempt?

We have a fake event source already, just no hold control; I'll ask the developer to add explicit release rather than another delay setting.

Assert the new attempt stays pending at that exact point.

The event has an attempt reference, Kai, but our handler matches the plate label; that is why the late result finds the newly opened row.

Also send it twice. Matching and duplicate counting can fail separately.

Then send the genuine new result. Rejecting everything must not pass the test.

Keep the stored result and displayed state as separate assertions, so the test can show which one changes incorrectly.

The queued sequence now fails reliably on the old handler, with display and stored-row failures shown separately; the replacement handler is not finished, but at least I can demonstrate the problem without narrating excuses for the timing.

Who keeps that test with the handover?

The developer will include it with the maintained test set, and I will use the same sequence in the tutor notes; restart identity rules still need their own case, so this isn't the whole recovery test.