It's not an odd request, and Sean answered a similar question a while ago.  Here is the start of the thread from his answer:

On Tue, Jun 18, 2013 at 1:44 PM, Chris Hollander <[log in to unmask]> wrote:
I'm not sure if this is possible right now or not (I haven't tried), but it would be interesting, and useful, to schedule agents according to a distribution. Right now when you schedule repeating, you tell it the start time and an increment, but what if that increment could be a probability distribution?

The motivation for this comes from the need to have agents act according to a Poisson process.

I think the way to do it at the moment would be to manually reschedule the agent at the end of each step... and to be honest, I'm not sure if a general solution in the form of an overloaded schedule repeating is even a good idea.