Postmortems

Blameless write-ups for incidents that consumed meaningful error budget. No fictional incidents are published here — only the template and mapping from the live board.

How /status/ maps here

Each open or resolved incident on the board includes a reference (INC-####), severity, service id, title, and a timeline (trigger, update, escalate, resolve) with UTC timestamps and Cloudflare colo.

Summary ← incident title + kind (down, degraded, burn)
Impact ← error-budget minutes burned (tracked when resolved) + board SLO window
Timeline ← incident timeline JSON rendered on the status page
Root cause ← written after investigation (not auto-generated)
Action items ← follow-ups with owners

Resolved incidents roll off the board after retention (~90 days) but postmortems may stay linked from handoff notes.

Template (copy for the next review)

INC-____ SEV_ · Title from incident

Summary

One paragraph: what happened, who was affected (readers of the public RPC board), and when (UTC).

Impact

Error budget consumed: ___ minutes of the 216-minute monthly budget (99.5% / 30d). Note any SEV1/2 pages sent.

Timeline

YYYY-MM-DDTHH:MM:SS.000Z · trigger · First automated or human signal
… · update · Mitigation step
… · resolve · Board green again

Root cause

Technical cause without blame. Include external provider links if relevant.

What went well

Detection (probe, burn alert, health check)
Communication

Action items

[ ] Owner — measurable follow-up

Status board Runbook