another bezel, another travelling scratch

No. Quality checked it and found no mark there. There's a handling rub near an edge, so we're keeping that separate from the bright line

Good, two different things to label. Don't let the real rub lend credibility to every bright stripe in the image.

Does the rub meet the product's rejection criteria? Its presence alone does not establish that this is a reject sample, and you will need that decision before using it to tune detection.

And what changed between the views? Part seating, light position, camera exposure? Someone will need those details to reproduce the stripe.

Quality accepts the rub. Camera and light stayed fixed; the bezel was reseated by hand. We've removed the scratch label and asked the fixture person to check the seating variation

Keep a genuinely rejected bezel in the next comparison too. Otherwise it's easy to make everything look consistent by hiding the surface detail you needed.