Hi Paul,
I used to use an AO-8 and ST-8 (and I wrote the original Maxim driver for the AO7 which we tested with an ST-7). For the really fast AO-7 (40Hz+) with the very small KAF402 the improvement was evident. For the slower AO8 (20Hz max) plus ST8 the device still had an impact (especially when not using a PME!).
I am more skeptical now of the results with larger chips. A couple of years ago I was gung-ho to make a large AO unit, or somehow fit the SX unit into my image train. But having read Allan Holmes last whitepaper carefully, the correlation between guide star wander and largish main chip image wander is not great. In fact I think he acknowledges that multi-star guiding may do just as well as tip-tilt AO for large chips, since only the low-frequency disturbances are correlated over many arc-minutes. Multi-star guiding averages out the high-frequency wander and you're left with corrections based on slow moving waves. In fact demonstrations that multi-star guiding works better than regular guiding is also the evidence that AO won't do much, if anything, for large CCDs.
But hey there is nothing stopping you from trying...what's two more adapters anyways
Best,
EB