I’ve seen cases where a functional block of an SOC got used one generation to the next without even seeing an issue.

And then suddenly, due to changing boundary conditions, a bug got triggered.

I’ve similarly seen cases where HW bugs were known but SW workarounds where calculates to reduce the chances of them manifesting themselves to be low enough. Until they suddenly didn’t.

The state space of HW is incredible, and it’s often not possible to foresee or understand the interactions between of SW and HW.