For a true random number generator, that's simply a matter of tuning the probability of random failure to be acceptable; e.g. the old FIPS 140-2 randomness test

For a true random number generator, that's simply a matter of tuning the probability of random failure to be acceptable; e.g. the old FIPS 140-2 randomness tests were calibrated to spuriously fail with probability of (IIRC) about one in a million. There are problems with testing a random source, but calibration can solve that one. (The bigger issue is that the "true" random going into your (PRNG) postprocessor should be whatever comes out of your best entropy-generating device, which typically isn't uniformly random; and that the data coming out of your postprocessor will look random to a black-box test even if the input contains basically no entropy.)