Metrics
Resources already emit into Effects per-process Metric registry (queues, processes, HTTP clients, runtime metrics). Two sinks read that registry:
| Sink | When |
|---|---|
| OTEL export | Professional stack — Sentry, Grafana, Honeycomb, alerting, retention |
| Telemetry | Custom in-app glass — fleet page, TUI, pm metrics — no sidecar |
Telemetry is the Hyperlink path: leaf snapshot / live, plus fleet inFlightByNode / fleetInFlight when the tag is meshed. Cardinality stays cheap (node / client / status labels; no per-entity metric ids). Per-entity current state still comes from that entitys own status / refs — not from Telemetry.
See the Telemetry guide.