Ingestion failures cause permanent trace data loss

Traces that receive 5xx error codes during ingestion are permanently lost and cannot be replayed. Only events that receive 2xx status codes are processed. During outages, most traces never reach ingestion endpoints.

Implement client-side retry logic with exponential backoff for ingestion failures. Consider implementing a durable queue or buffer on client SDKs to retain traces during temporary outages. Monitor ingestion error rates and trace acceptance rates to detect data loss early.