Set up observability

make observe builds scrape targets from the deployed fleet's engine IDs and metrics endpoints, then starts Prometheus 3.14.0 and Grafana 13.2.1. Prometheus collects router and engine metrics and evaluates alerts. Grafana displays them in the provisioned Narwhal Orchestrator dashboard.

Monitoring tasks

• Start and verify monitoring
• Access dashboards and isolate listeners
• Monitor a WSL2 development fleet
• Inspect GPU telemetry, alerts, and recovery