Marianne Oelund (a technically savvy engineer/photographer) make an interesting comment dpreview concerning the differences in how diffraction affects the D800 and D800E. I presume this difference is due to the lack of a low pass filter. When the Airy disc is already smaller than the pixel, making it even smaller by increasing the aperture has diminishing returns.
Assuming a perfect aberration free lens, diffraction will increase proportionally with the F-number as the aperture gets narrower. However, wide open the diffraction pattern is very small. In fact it is smaller than a single sensel. Therefore it cannot be resolved and has no meaningfull impact on resolution, because the sampling aperture of the sensel is the limiting factor.
When the diffraction pattern gets large enough by using narrower apertures, it's effect on the per-pixel contrast will become visible. The micro-detail gets lost because of the reduced contrast between pixels (they are contaminated with optical signals related to its neighboring pixels).
When we compare to a different sensor
with smaller sensels, the per-pixel contrast will be affected earlier when we use ever smaller apertures on the smaller sensel pitch version, however, we do get a more accurate (over-)sampling of the diffraction pattern. So, we lose contrast and detail per pixel (we can produce larger output at the same PPI, but lose detail), but for same size output we gain some resolution because we can use a higher PPI (and because the diffraction pattern is oversampled, we may be able to restore even more resolution by deconvolution sharpening).
So the whole debate is between larger size output (with lower per capture pixel contrast), and same size output (with higher per output size pixel contrast), which isn't much of a debate since they are different things.
Marianne Oelund is correct, when the diffraction pattern gets smaller than a single sensel (by using a wider aperture, and in the absence of residual lens aberrations
) then there is little resolution gain to be expected.
In the same thread, Bobn2 posted a graph showing how the f-number at which diffraction starts to become visible has nothing at all to do with the pixel size, which is somewhat at variance with Marianne's post.
I would have to read about the source data for that graph, before I can interpret what it tells. Maybe it becomes clear when I've had the time to read that whole thread.