You know, as I read this, one possible solution for his need for an ultra-fast pan/tilt mirror would be to use "galvanometers", but not hacked speakers- I would think the best solution would be pivoting the mirror with a selenoid vs. a spring in each axis.
Selenoids and springs are available for cheap money, and not too much is going to be faster without huge expense... PWM driving the selenoids would give you very precise control...
Since the device's purpose is aiming, calibration isn't needed- even if the resilience of the springs or the pull of the selenoid (brownout?) changes, you're correcting it on the other end. You could easily wind the selenoid coils yourself, and even use rubber bands (or even a couterweight, if you want to be downright medeival) instead of springs. The whole idea is that the selenoids are constantly pulling against an opposing force- pick your own poison.
I would think a design like this would be a LOT faster than any servo.. and potentially more accurate if you tweaked it right...
Last but not least, I'd been wondering how hard it would be to play around with one of those DLP chips they used for rear-projection TV's for a while there... all they are is aimable mirror arrays..