If a process crashes while holding a robust mutex, it’s possible to recover from such a situation.
But that's the whole point of poisoning too. If a different thread crashed and the mutex is poisoned, you can recover in a different thread. Most applications choose not to recover and chooses to just panic again, but you can choose to recover.
So basically the problem is bus factor = 1 and that one person is often away from the hardware and not able to remote in.
This has been happening for a long time it sounds like? This is very unfortunate but really could've been avoided with some planning it feels like.