I have worked at a company where every 500 error is a reason to go check the logs and try to reproduce... We would literally page an engineer.
I've also worked at a company where "as long as the error rate stays under 0.1%, it's fine". As long as each app server doesn't have more than 1 crash per week it's fine.
I can tell you that engineers at the first company end up finding all kinds of super rare bugs, usually in the OS, platform, malloc library, load balancer, etc. They then commit fixes for those bugs which ends up helping the latter company...
I've also worked at a company where "as long as the error rate stays under 0.1%, it's fine". As long as each app server doesn't have more than 1 crash per week it's fine.
I can tell you that engineers at the first company end up finding all kinds of super rare bugs, usually in the OS, platform, malloc library, load balancer, etc. They then commit fixes for those bugs which ends up helping the latter company...