application/problem+json payloads for errors.
Retry matrix
| Status | Meaning | What you should do |
|---|---|---|
400 | malformed request or unsupported combination | fix the request before retrying |
401 / 403 | authentication or permission problem | rotate credentials or fix access configuration |
404 | resource not found | verify the ID and tenant context on your side |
409 | duplicate or conflicting state | reconcile with the existing resource instead of retrying blindly |
422 | validation failure | inspect field-level errors and correct payload data |
5xx | transient server problem | retry with backoff and reconcile if the write outcome is uncertain |
Problem Details example
Practical retry policy
- Never retry
4xxwrite errors automatically except after explicit reconciliation logic. - Back off and retry
5xxresponses. - If you lose the response to a write, fetch the resource before creating a new one.
- Keep request IDs and resource IDs in logs for support escalation.

