Got lost in there somewhere. I'm saying the image data elements, as opposed to the metadata, are supposed to be identical. Which I think is what you are saying in the second paragraph.The data in the FITS file is not a bit to bit copy of the data in the ARW file, they are very similar but not identical. So these are different frames.
However, you are absolutely right: from the FITS file, crop 32 pixels from the right margin and you get the same image region as in the ARW file when opened with PixInsight.
Bernd
By the way, cropping is probably the RIGHT answer for this sensor. I haven't tried to tell precisely where this weird pattern starts, but the right side of an image (this is a very stretched master dark) is clearly a mess and not real data.
View attachment 8929