I suppose this topic could go under many of the categories already underway - Bridge reliability, Modeling, Prototype Failures, but it probably deserves its own topic.
Hubble Power Supply Failure
It probably goes down as one of the most expensive failures ever, and you can bet the calulated MTBF was about one million hours. There were probably more years of modeling that can be imagined.
Does anyone know what happened? Probably a well-guarded secret, but maybe a bridge? Or radiation?