{
  "schema_version": 1,
  "evidence_kind": "new actual-member experiment; not historical replay",
  "date": "2026-09-21",
  "run_id": "matrix-5920dd66b7",
  "verdict": "INCOMPLETE / FAIL",
  "victor_commit": "09f9894072c0c9dbe67cbb5e768e86a21a17f9fa",
  "source_clean_at_launch": true,
  "harness_sha256": "167d420367ac5ed1c96b985489b7d3eddff1312e0534a0e82b1c6452cca4c697",
  "private_evidence_sha256": "b0526882764dff9dcc1b82c7bc5512019ec1cea3196c0e6604c6fb08c634d392",
  "provider": "zai",
  "model": "glm-5.3",
  "case_contract_version": 2,
  "task_profile": "single-file",
  "cases": [
    {
      "case": "sequential",
      "passed": false,
      "failures": [
        "member_outcomes",
        "wire_routes_or_attempts",
        "missing_artifacts:first",
        "pytest:first",
        "semantic_oracle:first"
      ],
      "calls": 12
    },
    {
      "case": "parallel",
      "passed": false,
      "failures": [
        "wire_routes_or_attempts",
        "member_usage:executor_2f4373e7"
      ],
      "calls": 8
    },
    {
      "case": "hierarchical",
      "passed": false,
      "failures": [
        "wire_routes_or_attempts",
        "member_usage:executor_d6435919",
        "member_usage:executor_13750175"
      ],
      "calls": 9
    },
    {
      "case": "pipeline",
      "passed": false,
      "failures": [
        "wire_routes_or_attempts",
        "member_usage:executor_08bc2cfb",
        "member_usage:executor_f8d1a8c9"
      ],
      "calls": 16
    }
  ],
  "calls": 61,
  "http_status_counts": {
    "429": 26,
    "200": 35
  },
  "distinct_observed_member_sessions": 10,
  "accounting_observations": 61,
  "totals": {
    "tokens_in": 17100,
    "tokens_out": 2849,
    "cache_read_tokens": 58816,
    "cache_creation_tokens": 0
  },
  "buffered_gateway_deadline_seconds": 120,
  "shared_cache_cleared": false,
  "limitations": [
    "Independent numeric oracle checks a finite domain; it is not a hostile-code sandbox.",
    "Buffered calls only; no streaming or origin-cancellation acceptance.",
    "Reported cache counts do not establish backend executed reuse.",
    "Runtime identities and hardware acceptance must accompany this report.",
    "No tokenizer-equivalence or session-lease acceptance claim.",
    "Unsupported durable partial resume remains explicitly deferred.",
    "Single-file profile is distinct from standard two-file evidence; do not pool counts.",
    "Stopped after repeated HTTP 429s; four completed cases failed and the in-flight case was cancelled. No quota cause or model capability verdict established.",
    "Original detailed accounting failures remain in the preserved private evidence identified by SHA256; final wire rows and cancellation-time accounting are different snapshots."
  ],
  "accounting_failure_counts": {
    "http_failure": 28,
    "invalid_or_missing_usage": 10,
    "availability_coverage": 10,
    "c4_cache_observation": 30,
    "nonunique_request_join": 25,
    "ledger_settlement": 7,
    "deadline_reached": 2,
    "wire_ledger_call_count": 1
  },
  "accounting_passed": false
}