Broken context usually appears as separate traces where one distributed request should have a parent-child chain.

Checks

  1. Identify the outbound span in the calling service and the inbound span in the receiving service.
  2. Confirm that the caller injects context into the supported request or message carrier.
  3. Confirm that the receiver extracts context before creating its server or consumer span.
  4. Verify that intermediaries preserve the required propagation headers or message properties.
  5. Ensure both sides use the same approved propagator, or a deliberately configured compatible set.
  6. Check asynchronous handoffs for lost execution context and create links where parent-child semantics are not appropriate.

Do not log propagation headers in production diagnostics because they can contain operational identifiers. Reproduce with a controlled request and compare only sanitized trace and span identifiers. If a gateway rewrites headers, coordinate the correction with its owner rather than adding unrelated identifiers to application spans.