"...for most of that time, it was assumed that the problems had been fixed."
Isn't that, in principle, a testable assumption? I'd be interested to know if there is a testing discipline around such an assumption, I suppose it isn't the only time this has occurred.