Re: Boo, hiss
It's worth looking at how NASA do reliability, it's quite a departure from how most companies do it. (this is all from a couple of books I've read on it, no doubt someone will correct me). They decided a while back that if you're getting custom hardware built in small quantities and using it in applications that are really hard to simulate accurately you're going to have a hell of a time getting accurate MTBF figures. To get anything statistically significant would be impossible. So the alternative is to build things as well as possible and then all through the test phase take the approach that there are no unexplained failures. For example if you had a hard drive on a computer that failed, normally if it failed after a few thousand hours you'd just accept that these things happen and move on with your day. It would only be flagged as an issue if the same thing kept coming up again which is hardly meaningful if your sample space is 1 or even 10. The no unexplained failures technique means that when your HDD fails you then start digging and keep digging until you find the cause of failure and then fix the root cause. If you keep doing that process you end up eliminating the root cause of every known failure mode and hence increase reliability of your system. Expensive yes, but the only real way of ensuring reliability of a system that is effectively a prototype.
Taking this back to our current discussion and you're right on point. Yes there are failures, they get analysed, the root causes get eliminated and we end up with a safer rocket. What we should really be looking at is how the various organisations deal with failures and less about how many failures they have had.