PRNU & Stacking

As I am quite a bit out of my depth

if the Saturation is sensor-referred, why aren't you using all sensor-referred noise sources to calculate total noise? Alternatively why not go all ADC-input referred?

At what ISO does analog amplification stop and digital amplification begin for each camera you analyzed? No matter, we are mainly talking about the highlights here, right?

So as you stack more and more images the PRNU starts looking a lot less like noise and a lot more like an undesired signal - especially if pixel peeping at 100% or more. And in that case it no longer fits into the SNR equation.

Other than that your calculations appear correct to me (Bob, Emil?). As you stack images PRNU becomes the dominant unwanted artifact in the highlights very quickly. But, as I suggested earlier, it should not be too difficult to isolate the 2D 'PRNU' signal image and subtract it from the source. If your numbers are correct I think adding 8 K5 images at base ISO would provide a PRNU Signal to Shot Noise ratio of more than about 2:1, which should make it possible to isolate it. Once extracted I think you could apply it in the form of a 2D pixel efficiency matrix to virtually any image, no? But then again we are talking about the highlights. Do we really have a noise problem there?