And I wouldn't call diffraction a problem, per se - since every capture has some diffraction in the image plane, whether it appears in the print / on-screen or not.
I discovered the problem with the a7r5, it wasnt diffraction, it was slow sensor speed at 100ms using my extreme macro studio, the reason the a1mk2 was so sharp is the sensor speed is 8ms as im using electronic shutter with no black put viewing.
had me questioning as well. i knew my a6700 had some very minimal diffraction at 10x but couldnt figure out why the A1mk2 was so sharp even at 50 mpixels compared to the a7r5 at 60. i can also say that 8.5 stops of IBIS is total marketing crap, because it was no better than my 5.5 stop 7iv.
for anyone interested in the physics of diffraction, an interesting thing to observe is how the shadow of a building is not a sharp transition between sun and shade but that there is a narrow grey diffraction-induced zone between, This is the same effect that occurs between the edges of the aperture (or any border meeting the light while in transition in the lens, (hence the importance of large apertures for resolution)) With bellows or distance rings the recording medium is farther away and this will make the effect more visible since the rays/particles continue on their presumably linear path .
Good point, and Donald has mentioned his work with magnifiers e.g. 5x or 10x placed over the lens objective ... the question being:do magnifiers also affect the [focusing] distance of the aperture from the recording medium?
Illusion, exacerbated by upsampling methods that create pixelation (Nearest Neighbor and variants/hybrids). Instead of looking at how "sharp" details are on an insect, try both sensors with the same lens on text of variable size, like a document that scales font size by 0.8x with each subsequent line of text, printed and photographed from a distance. I don't even need to do this to know the outcome - there 60MP will record more lines, legibly, while the 33MP will give sharper results that are illegible gobbly-gook.