They don't have to thermally stabilize the system! They're the government, they can afford a mirror made of zero temp-co ZerodurTM, Cer-VitTM, or ULE. (It's just tax money, don't you know, there's always more where that came from!)
From the blurring around the orbiter tail in the photo of the silhouette, you can calculate the approximate effective shutter speed of the photo as being about 1 ms. This is a resolution limiting problem for that USAF system. IMO, they need something on the order of 1-10us and that might raise a small problem for the CCD manufacturers.
I sure hope that the low pixel resolution of the posted picture above is an artifact of the jpeg copies they made for public distribution and not an indication of the actual pixel resolution of the CCD they used.
Boot