First, thank you for responding in a manner that invites a response, rather than demands refutation.
I understand your perspective, and appreciate your recognition of my own. That said, if you think I'm putting the situation in an unfair light, I think you're downplaying it at least as much.
In my eyes, this is no mere "bug"; it's an abject process failure. As a reply to another of my comments in this discussion suggests, this is more on the level of, "Oops, we forgot to renew our domain name...", than it is, "Gosh, we didn't validate the pointer returned by the frobnitz function, when the whoozle isn't initialized yet..."
Dealing with expiring certificates before they expire is covered in like the second week of Certificate Management 101, as it were. If it's necessary to stick an intermediate cert in there, then it's doubly so to keep it current.
> The basis is that this is a single instance of a failure over the course of the features' lifetime, for a feature that has existed for absolutely ages.
The plural of "anecdote" isn't "data", but an existence proof is an existence proof. That the problem has gone from zero occurrences to one, no matter over what period, literally makes it infinitely more likely to recur, if you want to be that reductive...