I just sent a pull request that fixes a problem with the binomial random
variable generator.

In two places the old code used the stream of random reals generated by
the default (random-real) procedure instead of the stream specific to
the value of (current-random-source).

This was an error, in that binomial random variables generated for
different values of (current-random-source) would have been reusing the
same default stream of random reals for some purposes.

I suppose this could have been detected by a test, but it would be
difficult to devise such a test.

Brad