Comment by gregmac

> Hmm, I might be desensitivised from too much programming in erlang. It's implied that your program will encounter bugs or strange data and parts WILL be restarted, better account for that and plan on what to do on restart of each small part at the start of writing your program.

That's different from having software that can get into an unknown broken state, where physically power cycling it is the only fix.

It's better to crash and ensure there's a process that automatically restarts. Or use a watchdog process. Just breaking, to the point companies build external systems to reboot them, is just bad.