Running the tests again only works if the false positive rate is just due to noise. I.e. if the occurrence of a false positive is uncorrelated between re-tests.

Running the tests again only works if the false positive rate is just due to noise. I.e. if the occurrence of a false positive is uncorrelated between re-tests. I guess most real false positives aren't due to noise, but due to the test triggering on something unrelated and rare. In that case re-testing will yield the same result.