Give requests and jobs stable references
Carry a safe correlation or incident reference across HTTP requests, queued work and provider adapters so operators can follow one business action through the system.
Watch saturation and latency, not only errors
Track request latency, database time, slow queries, worker throughput, queue age, retry rates and provider error classes. An error-free queue that is hours behind is still unhealthy.
Structure events and redact secrets
Prefer named fields over unstructured text. Never record credentials, raw authorization headers or payment secrets. Hash or truncate identifiers when full values are unnecessary.
Alert on actionable boundaries
Alerts should tell an operator what boundary is failing and what business flow is at risk. Separate symptom alerts from root-cause evidence to reduce noise.
Use these patterns to structure a Thonem project, then verify the exact routes, providers, secrets, queues and infrastructure of the deployed environment before changing production.