{
  "example_kind": "original synthetic manual planning case",
  "slug": "flaky-check-attempt-history",
  "title": "Retain failed attempts after a retry passes",
  "scenario": "Synthetic planning case: A supposedly repeatable external check runs three times on the same declared artifact and setup: fail, fail, pass. A report keeps only the final pass.",
  "layout": "evaluation",
  "dataset": {
    "heading": "Inspect the invented case records",
    "headers": [
      "Attempt",
      "Declared artifact/setup",
      "Observed result"
    ],
    "rows": [
      [
        "1",
        "Same declared inputs",
        "Fail"
      ],
      [
        "2",
        "Same declared inputs",
        "Fail"
      ],
      [
        "3",
        "Same declared inputs",
        "Pass"
      ]
    ]
  },
  "derivation": "The recorded attempts include 2/3 failures and 1/3 passes. Those descriptive counts do not estimate a population flake probability or explain the cause.",
  "result": "Retain two failures and one pass with the artifact/setup context. The last result does not establish stable success or justify erasing contradictory attempts.",
  "boundary": "The guide preserves attempt instability, distinct from counting multiple reviews per accepted proof.",
  "limitations": "Three invented attempts cannot establish a calibrated failure rate. The claimed sameness of conditions requires independent evidence.",
  "sources": [
    "original"
  ],
  "tasks": [
    [
      "Preserve all attempts",
      "Test recorder",
      "The two failures remain accessible with the third pass."
    ],
    [
      "Check the sameness claim",
      "Investigator",
      "Artifact, setup and unrecorded conditions are examined before calling inputs identical."
    ],
    [
      "Resolve repeatability evidence",
      "Acceptance reviewer",
      "Any waiver or investigation is explicit; a final pass alone does not satisfy a repeatability requirement."
    ]
  ],
  "faqs": [
    [
      "Does the log prove nondeterministic software?",
      "No. Unrecorded setup changes, timing or evidence errors may explain the disagreement."
    ],
    [
      "Should every intermittent failure block everything?",
      "Apply the actual acceptance policy; this worksheet preserves evidence and does not supply a universal release rule."
    ]
  ],
  "method_references": [
    {
      "id": "original",
      "title": "Original worked-case definitions",
      "scope": "Definitions, policy choices, records and calculations are authored for this worksheet. No external standard, statistical validation or live product measurement is claimed."
    }
  ]
}