That's besides the point. Of course they need a test fleet. But in the absence of that, there's a very real chance that the existing bug triggered on customer machines but not their validator. This thread is speculating on the reason why their existing validation didn't catch this issue.