An easy path towards supporting DESC's desire to look at pre-interpolation pixels would be to save a mostly-zeros MaskedImage with those values as a separate intermediate data product somewhere in processCcd.py. Exactly what values to save (as a diff? with or without background? with or without a mask?) is still slightly TBD.
Should come with a Python snippet to combine this image with a calexp and/or postISRCCD, ideally with options to control what to re-insert by mask plane.
Storing these as HeavyFootprints with the Exposure may ultimately be a better solution to this problem, but is sufficiently harder to implement that we should at least do this first, and evaluate whether we need that after we've tried it out.