Your work on Pixel-shift with the Panasonic S1R and the thread that it generated is a very valuable addition to the discussion, the only reason that I didn’t mention it here was because the Panasonic ‘only’ does 8 shot so increases the resolution at the same time, it would have been very interesting to see the result had it been able to do 4-shot, and in-camera processing is such an advantage on the workflow (and file size) front.
Reading it all again I see that I’ve done Mark D Segal a disservice, he did share his test of 4-shot mode and found that it didn’t give him an advantage, at least in terms of the resolution of the detail in the dye clouds. I suppose the conclusion is that the processing of Bayer images is so good that any that theoretical difference that 4-shot pixel-shift might give is extremely difficult to demonstrate.
https://forums.negativelabpro.com/t/pixel-shift-vs-regular-capture-quick-comparison-w-s1r/8358/424
That also might mean that 4-shot is not seen to be a useful sales feature for manufacturers, and possibly why Panasonic chose not to offer it at all. That’s also probably why the Canon R5 MkII has moved from pixel-shift to AI interpolation. AI is such a marketing buzz word now and there’s none of that difficult science to have to explain, never mind that the actual results are shown to be inferior.