Ha, I think you have got me. I can't tell the difference between 1,2,5 and 6. 3 and 4 seem to have a hair less detail than the other 4 but seem the same as each other.
If you asked for a guess I'd say there are only two images 1,2,5,6 are the same and from a 5DSr, 3,4 are from a 5DS.
Picture numbers
1 2
3 4
5 6
Oooops....There were meant to be 3 images, not 6 ! I've uploaded them twice somehow. I bet you were cursing that I'd asked you to look at six identical images !
Thanks for taking the time to look and it's interesting what you have said; there are actually three different images, 1&2, 3&4, 5&6.
All are 100% crop.
1&2 are the 5Dsr, converted in ACR with zero noise reduction and zero sharpening.
3&4 are the 5Ds, converted in ACR with zero noise reduction but a USM of 100% 0.3 radius.
5&6 are also the 5Ds, converted in ACR with zero noise reduction and the same USM of 100%, 0.3 pixel, but then faded back 50%.
To my eye even the 100% of 0.3 in 3&4 had given an obvious sharpening to the 5Ds even though this is a very small amount of sharpening, and I guess this is why you thought it was the 5Ds - I had sharpened it to match the 5Dsr, and you were right, perhaps some detail was damaged.
The interesting bit is 5&6: once you fade back the tiny bit of sharpening on the 5Ds it is impossible to tell the difference between the s and the sr in this image. So your conclusion without knowing the differences is the same as the one I came to when I do know.
Even with the 5Ds it is possible to get moire if the pattern frequency matches the sensor, so personally I just do not think that it's worth doing away with the AA filter, even on a high res sensor.
Attached is a frame from a vertical pano, and all the windows have moire. It can be dealt with, but it's there to begin with. Incidentally I also shot the same scene on the M5 and it is the same.