It is not the probability of failure that is important, it is the probability of those failures occuring in 1 machine,...
I don't think you understand. A priori, given the chance of one drive failing inside a year as 1/100, the chances of two successive drives failing on you is 1/10000. But once the first drive has failed, the chances of the second failing inside a year is still 1/100, not 1/10000.
Granted, multiple orders are more likely to come from different batches ... but here's something interesting from Google - http://labs.google.com/papers/disk_failures.pdfespecially if the 2 drives in question are supplied from different batches and probably different factories, unlike the "4 ordered and 2 failed" occurance mentioned in your response.
Google did a survey on its disk failure rate and found a 2% failure rate in the first year, but 8-9% failures in years 2-3. Drive temperature was not a statistically significant factor. If Apple's computer HDDs fail at 2% a year, then 200,000 users who bought in the last year are going to see failures. That's a lot.