My thoughts exactly, Ken about sustained g, F-22 and higher performance legacy (and future?) aircraft.
I was involved in a study about off loading front line training to simpler, less expensive aircraft. A couple of interesting things came out of it - apart from the problems of replicating up to date front line software, displays and (in some cases) simulating sensors and systems.
First was a risk of negative training where the trainer did not exactly match the actual aircraft, mainly in terms of handling or performance. Whilst this was found to be less of a problem for more experienced aircrew, there was a risk of regression for newer pilots.
The other thing was from an engineering standpoint, whereby the entire logistics and maintenance empire became geared up (and established) for a lower flying rate, which was then overstretched to support the increased flying rate required for excercises and ops where the trainer could no longer substitute for the real thing.
When you need your trainer to emulate more than one type, of course, the complexity of the task becomes huge.