I'd want the proposed holder involved early. Small shifts might be intended variation or poor seating. That changes what the imaging arrangement is expected to tolerate, and I wouldn't choose the easiest photograph as the nominal pose just because it looks calm.

Parts are retained, but the image filenames don't identify which faceplate is which. Can we recover that link?

Only if there is independent evidence, such as a capture sheet or visible identifier. Comparing the supposed scratches to decide which part it was could bake the disputed classification into the identity link as well.

Could new labelled captures be less work than reconstructing an uncertain old sequence?

Yes, though I wouldn't throw the old sequence away. It still shows the reflection problem that started this. The new set can have controlled identities and the old one can remain clearly unsuitable for per-part acceptance comparisons.

Carla has not said whether a capture sheet exists. Check for it before declaring the original sequence untraceable; an absent filename field is not proof that the identity record was never kept.

I haven't checked the demonstrator's notebook yet. Asking now; no decision to discard the images.

Fair correction, Harish. I jumped from bad filenames to a missing link. A separate record could recover it without reading identity from the disputed marks.

Keep recovered identity and physical classification as separate steps, whichever route you take.

And keep the holder question open beside them. Knowing exactly which plate is shown won't tell us whether the changed presentation was within the intended seating condition. I don't want the new reference set to silently normalise a fixture issue.

Notebook found. It identifies the first faceplate, then says swapped without naming the second. First sequence recoverable, remainder uncertain.

Then recover the first sequence and leave the rest at their actual confidence level. You don't need to make one decision for the whole folder. Does the first sequence include both the bright and less bright view of that identified plate?

Yes, both views. That plate is going to quality with the capture notes; holder seating limits still need confirmation.

So the new captures can target the missing identities rather than replace everything. No physical verdict yet, though.