It is important to note that the Repetition Count Test described in Section 8.6 is not a very powerful health test. It is able to detect only catastrophic failures of an entropy source. For example, a noise source evaluated at 8 bits of min-entropy per sample has a cutoff value of 5 repetitions to ensure a false-positive rate of approximately once per four billion samples generated. If that noise source somehow failed to the point that it was providing only 6 bits of min-entropy per sample, how many samples would be expected to be needed before the Repetition Count Test would notice the problem?