Hmm, @bashtage the von-mieses tests fails in one of the runs here:

    @pytest.mark.parametrize("mu", [-7., -np.pi, -3.1, np.pi, 3.2])
5690
    @pytest.mark.parametrize("kappa", [1e-9, 1e-6, 1, 1e3, 1e15])
    def test_vonmises_large_kappa_range(self, mu, kappa):
        r = random.vonmises(mu, kappa, 50)
>       assert_(np.all(r > -np.pi) and np.all(r <= np.pi))
E       AssertionError

kappa      = 1000.0
mu         = 3.141592653589793

I guess this can probably be assumed to be rounding errors in the RNG that are exceedingly unlikely and can't be avoided (i.e. the test is potentially stricter than we can guarantee)? Or is this worrying?

That random looks like it isn't seeded. It should get a seed, or these will continue to occur with some tiny probability in the future.

Thanks Bas.