# Expected Diagnosis: ttl-row-processing-in-premium-cluster-during-sql-tail-latency-incident-window

This ad-hoc case checks whether the agent can recognize that TTL was actually
active during the latest completed Beijing `11:00-11:05` window on the same
`the primary premium cluster` cluster used by the other ad-hoc evidence cases.

The runtime hook resolves the latest completed Beijing `11:00-11:05` window at
execution time and computes the live reference from the real TiDB TTL counters:

- `tidb_server_ttl_insert_rows`
- `tidb_server_ttl_processed_expired_rows`

The expected agent response should:

- query the real TTL counters instead of inferring from slow-query or generic
  log absence
- emit exactly one JSON code block with the required fields:
  `cluster_id`, `window`, `evidence_sufficient`, `ttl_activity_present`,
  `signals`
- keep `signals` limited to:
  `ttl_insert_rows_delta`, `ttl_processed_expired_rows_delta`
- return the same cluster id, window, boolean status, and counter deltas as the
  runtime reference