Just a heads-up. I've checked in an enhancement to RANDOM that makes random much faster for numbers between 1 and 2^32 and also gets rid of the small bias in RANDOM.[1]
This change is incompatible with previous releases since the sequence of random integers from the same starting state will be different from before. Ray Footnotes: [1] RANDOM was computing random integers using (rem <random 32-bit int> val). This has a small bias in it.