Remote incident handoffs without duplicate noise

Incidents fail in the seams: the five minutes when Commander A steps off and Commander B has not read the latest thread. We standardize a handoff block with timestamp, customer impact sentence, active experiments, and explicit “do not touch” zones. Anything else waits for the post-incident doc.

Voice channels stay small. We route observers to a read-only stream so primary channels stay quiet enough for decisions. That rule sounds strict until you hear a tired engineer repeat the same mitigation three times because newcomers asked overlapping questions.

We end every major incident with a five-line summary aimed at people who were asleep. If those lines cannot be written without jargon, the incident is not actually over.