From Lurk:

It was generally believed that the probability of reactor failure is 1 time in a million years. Practice has shown that the first million years have already passed, and the probability is calculated correctly.