Bring a Docker test rig to a fresh, reachable gateway.

Promote `project` from gateway profile `profile_a` to `profile_b` via ign's project_sync. Refuses when the two profiles are already identical, and requires confirm: true to actually push (project_sync is destructive). The second project_diff is a verification, not a report: resources still added or changed on `profile_b` mean the sync did not land, and the result is `verification_failed`. Resources only left over on `profile_b` are reported as `pending_removals`, since removing them needs delete: true.

Full gateway health read in one call: status, license, redundancy, GAN, connections, modules, doctor. Verdict is `down` when `status` itself fails, `degraded` when any later step fails, when any module reports a state outside {ACTIVE, RUNNING}, when any database connection has `enabled: false`, or when any doctor check reports status `Fail`, and `healthy` otherwise.

Search active alarms by priority and source patterns, return sorted by severity.

Daily gateway health check runbook.

Tear down and bring up the active rig, then wait for the gateway: rig_down, rig_up, wait_gateway, status. Requires confirm: true. ign's `rig_down` verb is not itself guarded (unlike rig_reset, rig_restore, and rig_trial_reset), so this composite enforces its own confirmation gate before making any ign call, rather than forwarding `confirm` to rig_down.