Core answer
Start here. This is the direct spoken answer to practice first.
Why this question matters
Asynchronous work often breaks the neat request trace. This question checks whether you can keep enough context across queues without leaking sensitive data.
I start with one failed job and compare the producer span, message metadata, consumer entry point, and first worker log. That identifies whether context was never injected, was dropped during serialization or broker mapping, or was ignored when the worker started a new root trace. I keep the durable job and message identifiers beside the trace data so I can still locate the exact delivery even when tracing is incomplete.