tmax is too large for an in-auction call
tmax exceeds 1000ms. The extension point runs inside an auction the exchange is already timing out, so a budget this large usually means seconds were sent as milliseconds.
What it means
tmax is above one second. ARTF's reference samples run at 150ms, and the extension point is called inside an auction that the exchange itself is timing out in the low hundreds of milliseconds.
Why it matters
The common cause is a unit slip: a 30-second service timeout copied into a field that counts milliseconds. The agent then sizes its work for a budget it will never actually be given, and the orchestrator abandons the call long before the answer arrives.
How to fix it
Set tmax from the auction's own remaining budget, in milliseconds, minus expected network latency. If the value really is intended, nothing is broken; the warning marks it as worth a second look.
Example that trips it
{ "id": "ep-1", "tmax": 30000, "lifecycle": "LIFECYCLE_PUBLISHER_BID_REQUEST", "bid_request": { "id": "a-1" } }30 seconds where 30 milliseconds or 300 was meant.
Check your payload
Paste a bid request or response into the tester to see whether this code fires against it, or gate on the id in CI with the CLI: the id is stable, wording is not.
Back to the full diagnostic code reference.