NGTM-1R: Thanks, I'll ask around here and see if anyone knows anything about it.
Why does the star brightness affect the adaptive optics result? Is this because of the longer exposure time required, thus the mirror has to deform more to compensate the changes along the path length? I suppose the Strehl ratio is easier to reach with the infrared wavelengths due to diffraction limit being larger, but I'm not sure about the seeing and it's effects in the infrared waveband. Also, I have been a little bit suspicious of what happens with the deformable mirrors with regarding the diffraction spot, it should not look like an Airy function any more if the reflective surface is not spherical.
For AO to be effective, the exposure time for the guide star must be relatively fixed, as you must update as least several hundred times per second. As you go for fainter guide stars, the diffraction pattern is less and less well sampled; for full correction, you need an extremely well sampled image, thus restricting you to very bright stars (for an 8-meter, 8th mag is insano bright).
Seeing comes in because the worse the seeing, the stronger the correction required. If the atmosphere is giving you a 1" disk, this is obviously harder to fully shrink than a 1/2" disk. So, because we're restricted in how long of an image we can take and be effective, worse seeing means lower Strehl, etc. This will continue to improve as CCDs get better; the paper I linked to talks about LLL CCDs a little. Those should be designed and prototyped in the next decade or so. ELT class instruments like TMT and, well, the E-ELT, might be among the first to use them.
Visual wavelengths are harder to do than IR for two reasons: first, perfect Airy disks are smaller in the optical, and second, the index of refraction of the atmosphere is higher than in the infrared, so the image intrinsically spreads out more before it even gets to the telescope. Even so, 20% Strehl is actually a very impressive result, since open-loop (i.e., no AO) operations on large telescopes typically net you 1% Strehl at best, even on nights of excellent seeing.
To your last point, the diffraction pattern for an image is the Fourier transform of the aperture function (we're waaaaay in the far-field limit here). Since the aperture function is still almost an evenly illuminated circle, the diffraction pattern is an Airy disk for a point source. Furthermore, the reflection surfaces are never spherical anyway; they're usually hyperboloids, though their perimeters are always circular if possible. Segmented mirrors like those on Keck and some other large telescopes introduce whole new problems; their PSFs are usually pretty funky and more difficult to deal with.
This suggests that there are physical limits on ground based telescopes that do not allow them to achieve the equal technology level space based telescopes, ever (such a dangerous word for a Physicist, ain't it?). I wasn't aware of that AO telescopes may only achieve 20 % Strehl ratio at visual wavelengths; that's actually pretty bad. I would love to see the MTF curves of those, though. It's hard to judge between Strehl ratio and MTF, Strehl ratio being relatively stringent.
Oh, ground based telescopes will probably catch up eventually, but they aren't quite there yet. Pyramid wavefront sensors are, for now, the way of the future, and they'll continue to get better. It wouldn't surprise me if the ELTs are left Hubble in their dust by the time they were done. Since their mirrors will be ten times the diameter of Hubble's, even 20% Strehl means they beat Hubble's angular resolution. The era of Hubble-type visible light missions is probably over. Such IR, microwave, X-ray, and gamma-ray space telescopes are going to be launched for a while yet, though, until we can put them on the far side of the Moon or something.
For once it would be cool to pull of the complete system design for a telescope, considering all that affects the resolution (struts, sensor pixel pitch, manufacturing errors, etc.). For example, the secondary mirror struts tend to have funny effects on the MTF, though the astronomers do not usually complain about the loss of resolution because of them. Perhaps it is impossible to do away the struts, that would likely end up with resolution problem being switch to color reproduction problems.
We don't complain about the struts because their effect is small and unavoidable. The struts cover only a tiny portion of the field, and while they do introduce diffraction effects, those usually pale in comparison to those of the atmosphere and other factors.