Separate physical judgement from image visibility. Moving brightness can be a reflection, but it does not prove the plate is free of damage beneath it.

And link each capture to the plate, holder condition and lighting setup, because a trustworthy physical reference can still have an unexplained image

No purchase planned. The retained plates are identified and quality is judging them under the appearance requirement. We have enough setup information to repeat two example views, but not the whole original image folder.

Do the repeated views show the same required surface regions, or does one hide an edge while making the central streak less bright?

Fiona's question is important because a cleaner picture may just contain less inspection. I would put the required regions beside each candidate view, then ask whether the judged defects remain visible over the allowed presentation variation rather than choose whichever single image looks nicest.

One view does hide a narrow edge strip. We have excluded it as a complete inspection view. Quality judged one streaked plate acceptable and retained another with a real edge defect as a reject reference.

Could the edge have a separate view? I am asking about coverage, not assuming another camera is needed.

A separate view is an option to assess, provided the combined sequence covers the required regions consistently. It also needs an explicit incomplete-inspection state when a required view is missing; one good centre image must not become a whole-plate pass.

That missing-view state needs to be understandable in the handover too, because the relief operator may only see a decent-looking centre picture and wonder why the plate is held.

Do we have a repeatability result yet, or only a proposed second view?

Only a proposal from Oliver, as I read it. We shouldn't quietly add a completed camera sequence to Clara's evidence.

Still a proposal. Repeating the existing full-face view on the identified plates has shown the edge defect becoming hard to see at one presentation, while the acceptable streak changes brightness. No threshold choice resolves that coverage problem.

That supports investigating presentation and illumination. It does not yet say which one should change; keep the difficult presentation in the comparison rather than remove it because it spoils the result.

Is that difficult presentation within the proposed fixture's permitted variation, or an accidental seating state that should be treated separately?

If seating is uncertain, record that uncertainty with the view instead of assigning it a tidy angle from the nominal drawing.

Fixture reviewer found the difficult view came from a plate not fully seated. We are keeping it as an abnormal presentation example, not removing it. The intended seating range still needs its own comparison.

Can the capture record make that distinction visible without making the operator choose a defect label? They should be able to record uncertain seating without deciding whether the plate is scratched.

Yes, separate presentation status from the quality reference judgement. Have the responsible designer define what prevents an incomplete or invalid presentation from being reported as inspected.