Readout noise does not affect sharpness, it affects the SNR.
SNR is a relevant factor to sharpness, taking that to mean (fine) contrast resolving power. Readout noise can reduce the SNR and thereby the resolving power of the signal. Which does in the extremes affect sharpness (though probably not on ISO 100 images taken in controlled studio conditions. But you were saying that there was no influence on sharpness whatsoever, and that's just wrong.)
The area covered by the microlens/pixel will not affect sharpness, it will affect light gathering ability.
I was thinking about that. I was not entirely sure. However, I figured that it'd have an effect in more smoothly rendering small detail, when the pixel is a full square rather than when it is a tiny hole covered with non-light-gathering gaps in between (to take an extreme example). Obviously, light gathering ability is the major difference, and a sharpness increase will be tiny. But I'm tending to think that there will be SOME effect.
Compare a grid of small circles (the circles being light sensitive, the area in between the circles not) and a grid of squares (100% coverage pixels) and compare their resolving power. The grid of small circles will miss some light (coverage), which will in most situations also include contrast differences (which equals detail). Cropping out contrast differences can't have a positive effect on sharpness, I'd think.
Sensor well leakage only happens when a pixel has too much light in it.
I thought its effects became more significant as the wells filled up (perhaps exponentially so), but that it is still present below the threshold of a completely "full" well. If you're correct, then it indeed has no impact on sharpness or contrast resolving ability.
Whether the above are or are not possible factors, that still leaves the AA filter as well, of course.