Operate: incident-time answers for a modular feature

Operate is a new section built for the middle of an incident. Its hub answers nine questions in one line each, with the single command or link to use now, and seven pages carry the detail:

• What is running, and how do I roll or roll back? The deployment project’s last good production upload is the answer to both. Rollback is ordered by speed: flag off, retry the last good pipeline and hold production behind a manual play, then revert.
• Pod health. The restart query, the probe events, and how to reach LabKit’s probe endpoints.
• Safe rollout and feature flags. What the Kubernetes runtime does and does not offer, and why the flag is your canary.
• Which cell. On Runway, split by region and cluster. On Cells, the cell is the tenant label on the alert, and cell metrics never reach central Mimir.

Every page has an “On Cells” part. Where the per-cell path for a modular service is not ready, it says who owns it and links the tracking issue rather than inventing a procedure. The gaps page now tracks those holes.