When calculating the reliability of systems with multiple disks such as RAID arrays, it is conceptually obvious that, if one drive has a probability of failure of one in a thousand, the probability of two drives failing is one in a million because independent probabilities multiply. Why is this statement probably false?