Debug-Distributed-Lock-Expiry-V2

The model receives the prompt (and optional system message). The run uses scorer rubric_json_metrics with the JSON configuration below. Pass/fail and partial credit are determined entirely by that scorer against the model output; no human grading.

Return JSON only with keys diagnosis, fix, tests. A distributed lock is acquired with a short TTL; the lock holder crashes mid-operation after TTL expiry; a new process acquires the lock and overwrites state that the first process had only partially updated, leaving data in an inconsistent state. Identify the partial-write hazard and fix it.

{
  "metrics": {
    "repro": {
      "checks": [
        {
          "contains": [
            "TTL"
          ]
        },
        {
          "contains": [
            "partial update"
          ]
        },
        {
          "contains": [
            "expired"
          ]
        },
        {
          "contains": [
            "inconsistent"
          ]
        }
      ]
    },
    "hidden": {
      "checks": [
        {
          "contains": [
            "fencing token"
          ]
        },
        {
          "contains": [
            "version check"
          ]
        },
        {
          "contains": [
            "conditional write"
          ]
        }
      ]
    },
    "diagnose": {
      "checks": [
        {
          "contains": [
            "lock expiry"
          ]
        },
        {
          "contains": [
            "partial write"
          ]
        },
        {
          "contains": [
            "split-brain"
          ]
        }
      ]
    }
  }
}