{
  "id": "20261002T164150404672Z",
  "decided_at_utc": "2026-10-02T16:41:50.415128+00:00",
  "status": "continuation_stopped",
  "continuation_sha256": "439e844fd74b119eef354ba2ff367fb0039b7f6438c8e0c89270f0b16e29de46",
  "pre_continuation": {
    "responses.jsonl": {
      "sha256": "ea56e931c3509298d6f34ad7fd68c6dcc678cd41be34b50b1c1281541d113998",
      "bytes": 11831768
    },
    "attempts.jsonl": {
      "sha256": "63597db4441ebdbfe4758b4003cf1c6cb5a920a7eb0f029bc065ea944d802864",
      "bytes": 460797
    },
    "plan.json": {
      "sha256": "b6f47e7d5913a3be6c34e0ac8b59ee7e8e70953907a53a002818286c6887327f",
      "bytes": 1342472
    },
    "requests.jsonl": {
      "sha256": "eca30790f8e18d580b19662c3d61cee27f4f8de6db4a3c3b8ba31e45a4b23911",
      "bytes": 9149297
    }
  },
  "metadata_before_sha256": "537352319561e3368e72879891a89233dc103e1c820282ac7c9c31f9e2f16b96",
  "metadata_before_path": "benchmark/multimodel/runs/quality-20261002T155349573021Z/continuation-history/20261002T164150404672Z/metadata-before.json",
  "recovery_run_dir": "benchmark/multimodel/runs/probes-recovery-20261002T164150404672Z",
  "original_attempts": 3019,
  "original_invalid": 47,
  "unattempted_requests": 1766,
  "models": [
    {
      "label": "Qwen3.8 Flash",
      "id": "qwen/qwen3.8-flash",
      "provider": "alibaba",
      "provider_name": "Alibaba",
      "prices": {
        "prompt": "0.00000015",
        "completion": "0.00000047",
        "input_cache_read": "0.000000016",
        "input_cache_write": "0.0000002"
      },
      "reasoning": {
        "effort": "none"
      },
      "cache_control": "unsupported",
      "max_tokens": 128,
      "reasoning_mandatory": false,
      "protocol_difference": null
    }
  ],
  "workers": 16,
  "seed": 20261002,
  "total_quality_and_recovery_budget_usd": 5,
  "change": "After two valid synthetic service-health probes with observed zero reasoning, clear only the replayed historical consecutive-failure stop and streaks. Preserve cumulative invalid counts, every original attempt and bill, all unknown-bill reservations, request hashes, worker16 and the original live circuit. Dispatch only previously unattempted sequences. No benchmark output was selected for removal or retry.",
  "recovery_finished_at": "2026-10-02T16:41:52.652944+00:00",
  "recovery_budget_accounted_usd": 6.534e-05,
  "historical_recovery_budget_accounted_usd": 0,
  "recovery_calls": 2,
  "finished_at_utc": "2026-10-02T16:42:49.088229+00:00",
  "final_quality_calls": 3651,
  "final_quality_unknown_bills": 71
}