@regehr makes sense, agreed. OTOH theres another extreme opposite viewpoint on it which says that even if all projects start as pets they should evolve into cattle eventually. in cattle we dont want any SPoFs, for example, but also we should assume that anything that can go wrong in prod will eventually go wrong in prod. every possible edge case will happen, one day, perhaps at the worst possible time. thus, chaos engineering. the only way to know for sure if one's system can recover from X is to forcefully induce X and see what happens. granted killing all processes as root should not happen but... its possible. heck look at the flood of old Linux LPEs surfacing in recent months