The WarpCompare results look quite similar to the SafeClip ones, and that made me think I had misdiagnosed the behavior of the new configs around bright stars.
I had - the problem is that we have some pixels masked with INTRP in the single-epoch images with no additional bit set to indicate why they were interpolated. I think I'd call that a bug, and I've opened
DM-12985 to keep it separate from the coadd configuration.
Because the config change proposed by this ticket does make the bright stars on the coadd look a bit worse (because of those INTRP-only pixels), I'm inclined to block this merge on
DM-12985. If that takes too long to sort out, however, we may want to press forward with this one on the principle that avoiding overzealous INEXACT_PSF masking outweighs any damage done to saturated stars.