Because of shared memory most panics and seg faults in a worker process take down the entire server already (this wasn’t always the case, but not doing so was a bug).