Result type: SYNTHETIC · fake-adapter harness · not a provider benchmark or production result.
This page records the safe scope of the KD11 V20 WAIT_RESOURCE / WAIT_AUTH recovery exercise. It is a public summary, not raw logs or a live status feed.
Recorded scope
- WAIT_RESOURCE lanes simulated
- 1,000
- WAIT_AUTH lanes simulated
- 1,000
- Synthetic resume events reported
- 2,000
- Provider calls
- 0
- Provider tokens
- 0
- Production jobs
- 0
These are harness scope counts. They are not production job completions, user output, a measured recovery rate, elapsed-time performance or evidence that a real provider resumes any task. The source summary does not establish a per-job success rate or accepted artifact count, so those values remain UNKNOWN.
Interpretation
WAIT_RESOURCE represents a simulated capacity/provider wait; WAIT_AUTH is a separate simulated authentication wait. They should route through distinct recovery handling. A provider-free harness can examine a state-machine path without making a provider call, but it cannot establish real retry timing, credential refresh, useful output or quality.
No task titles, lane identifiers, session identifiers, prompts, private messages, credentials, local file paths or raw log text are included. The public-safe article copy is limited to the aggregate counts and interpretation above.
What is needed for a production conclusion
Record authorised production attempts with a durable job identifier, wait start/end, attempt reconciliation, checkpoint, provider usage, human handling and independent verifier outcome. Keep failed and unfinished jobs in the denominator. Until such a joined source exists, production recovery effectiveness remains UNKNOWN.
Related: recovery outside the model turn · WAIT_RESOURCE method · WAIT_AUTH handling · operating rules tool.