With practice CATs, the question pool is usually just a few hundred questions deep. I have discovered this after experimenting with the same Kaplan or Upstart Raising CATs multiple times. Basically, the same "difficult" questions will simply appear earlier in the exam after readjusting for a new performance level. Obviously this is not an accurate representation of a test takers true score; not to mention that many of the problems come directly from their textbook and online q-bank problems. Same deal w/power prep.
With the official ETS test on the other hand, given that there are a finite number of topics testable (hence its standardization), is it conceivable that there are an infinite number of unique question combinations (diregarding the experimentals) thereby making it a bottomless pool?