It's more subtle than that, and it doesn't require a repeating event. Suppose you have an event that combines both "rand" and another sexp using "and". The code must continually evaluate the two conditions until both are true.
The original implementation of "rand" had the sexp short-circuit its evaluation by replacing itself with the chosen random value. In effect, the random number was only picked once, and it remained consistent for every evaluation of the sexp. My original change caused "rand" to pick a new random number every time the sexp was evaluated. (This is what "rand-multiple" does now.)
I can't really think of an example to illustrate the result of this, but you can think of it as a sort of quantum uncertainty. The value of "rand" is unknown until it is locked in place when both halves of the "and" sexp become true at the same time. I forget the exact situation which clued me in to this difference, but it definitely exists as evidenced by the fact that TBP noticed it.

As an aside, "rand-multiple" is probably not the best name for the new sexp but it was the best I could think of at the time. Do you have any suggestions? "Quantum-rand", maybe?
