Operate WebhookCore

Signals to watch first

• active endpoint count
• delivery backlog
• retry volume
• replay volume
• failure reasons by status or timeout/network outcome

Useful surfaces

• GET /health
• GET /ready
• GET /metrics
• GET /v1/deliveries
• GET /v1/deliveries/:id/attempts
• POST /v1/deliveries/:id/replay

Good operational posture

• treat secret rotation as a normal operational action
• keep endpoint activation state visible
• forward x-request-id and traceparent into downstream logging
• protect /metrics through network policy or a reverse proxy

Failure modes worth rehearsing

• inactive endpoint: delivery creation should reject with 409
• downstream outage: retries should stay explicit and auditable
• timeout versus network error: attempt outcome should stay distinguishable
• replay: should preserve prior attempt history instead of overwriting it