And, then again . . .
If you do anything that modifies the relationship between the CFA and the pixel field, wouldn't this have a detrimental effect on the DeBayered data - one that could well require a second round of cosmetic correction later in processing?
Wouldn't a
better
approach be to perform all the calibration, deBayering, registration amd rejection-based integration steps first - and then to inspect the resultant MasterLight to see if cosmetic correction would have been necessary?
I don't know the answer because I don't use a DSLR and therefore have no experience as to why DLSR images would require cosmetic correction where other OSC cameras do not (for example, I have never needed to use this process).
Perhaps the only way to answer the question would be to pre-process the same set of raw data in two different ways, and then see what the outcome is?