Sorry, it's time to close this one.  The code is not buggy, it is documented as being a bound method and that is a perfectly acceptable python coding style to use bound methods as callables.  It's also been around for a *very* long without causing issues beyond Python's usual irritation with the number of arguments messages (that happened when Python became object oriented and that artifact still survives).

When it comes to the random module, people are performance sensitive and it would be a total waste to wrap these bound methods in another layer just to get different error message reporting.  Anyone whose Monte Carlo simulation run time doubles because of this would be rightfully irritated.