I took the dashboard total to the bench and found sixteen more checks than housings. Reviewed retries explain them. They do not explain why the FR10 report calls all ninety-eight finished parts.
We have item identities and disposition records. I want the accepted count corrected without erasing the failed attempts that tell us how much work this batch actually took.
Keep attempt workload as a separate measure. It is useful for planning, just not as the number of physical housings accepted. No need to throw away the busy part of the story.
Does the headline total also appear in the exported report? A corrected screen would be a rather small victory if purchasing still receives the old number.
Sarah, yes, all ninety-eight link to eighty-two housings. Bruno, export repeats the same inflated count. Current dispositions are seventy-six accepted, four held and two rejected, which is less exciting but actually adds up.
Use the current authorised disposition for each identified housing, with its attempts retained beneath it; then repeated delivery of a result should not produce another accepted housing.
And test a later change of disposition. I mean an authorised change, not a software guess. The count needs to reflect it without deleting why the earlier decision was made.
The revised offline report shows the three disposition groups and the separate ninety-eight attempts. Duplicate delivery leaves the item totals alone. Later disposition changes and restart are still on the maintainer's test list.
Quality owns that comparison against the disposition records. I have sent the actual housing list, not just the three numbers. Until they check it and the remaining tests pass, this is a test report.