Antoine, there does need to be repeatablity; there's no question about that.  The open question for me is how to offer that repeatability in the cleanest manner.

People use random.seed() for reproducible tests.  They need it to have their studies become independently validatable, etc.  Some people are pickling the state of the RNG and need to restart where they left off, etc.

Mark, thanks for the alternative formulation.  I'll take a look when I get a chance.

I've got it from here.