Source attribution must become complete without making customer trace volume unpredictable.
Incident investigation needs identity on rejected requests, while retained spans determine what customers pay to observe.
Who you’d be doing this for
“I had twenty rejected calls in one burst, and every trace stopped short of who sent them.”
Parisa Jafari · AI Application Engineer
Operates a production agent and uses traces to explain failed requests and token spend to platform and finance partners.
What is at stake
Rejected LLM requests lose their calling source during traffic spikes, blocking investigation and source-level cost analysis. You must weigh attribution coverage against false labels and permanent span-volume exposure.
Why it isn’t already fixed
Every obvious fix costs something else. That’s the part you’d have to decide.
- source coverage vs false attribution
- incident context vs predictable span volume
- gateway latency vs earlier metadata capture
- fast cutover vs verified production behavior
Why Datadog
LLM Observability connects agent traces with cost and production telemetry, so unassigned rejected requests break the investigation chain customers rely on.
Written with these in mind
Not your kind of problem? 40 more at Datadog, or browse every organization.
This is the setup. The work is inside.
Running it puts you in the room: the full situation and its constraints, stakeholders who push back in their own words, and the decisions that are yours to make. What you produce becomes a Day One Plan — work you can show someone instead of describing.