{
  "summary": "Exhaustive inventory: 11 distinct actual rig receipts = golden + four PASS positives + five analyzed negative controls + one additional startup-only negative attempt. Five analyzed negatives all have 23 post tags, zero complete/timely, no recovery, one replacement poison. Two retain outer CLEANUP_FAILED but predicate_status REGRESSION. Additional negative-1 is PRECONDITION, not replayable as a refresh regression. No predicates/analyzer, tests, builds, or live rig executed; only read-only retained-JSON arithmetic and SHA256.",
  "files": [
    {
      "path": ".spt/preserved/307/golden-35299506382/deployah-refresh-receipt.json",
      "description": "Golden; SHA256 8bc3d37b5a82c195173eefc615b688970776437992642c0a1bdb84adac49284d"
    },
    {
      "path": ".spt/preserved/307/gate-228/rig-1/observation/receipt.json",
      "description": "Positive; SHA256 e6ecad2bedbb02a45dbdbd72c0a0e1c00c1b1ba492a02cc2af0fcff5ce07e555"
    },
    {
      "path": ".spt/preserved/307/gate-228/rig-2/observation/receipt.json",
      "description": "Positive; SHA256 96a7954fbd2c2f4ed05624fa16b43a0f9c4b00b74fd09b58b8794ad5b143cc74"
    },
    {
      "path": ".spt/preserved/49-267-refresh-discriminator/windows/candidate-1/rig/observation/receipt.json",
      "description": "Positive; SHA256 e3198a10f04768b24c20a33127b7946efe323441c116ef2fb216207ce30ea0e4"
    },
    {
      "path": ".spt/preserved/49-267-refresh-discriminator/windows/candidate-2/rig/observation/receipt.json",
      "description": "Positive; SHA256 2199584194f9ca3b49a8104a4b431303b06e2d3b2a52c48800aa07c6b22915f2"
    },
    {
      "path": ".spt/preserved/307/hertz-w2-regression/negative-289e3cdf-1/observation/receipt.json",
      "description": "Additional startup-only negative attempt; SHA256 71b64063351b72db4413a102bfc4790daaabbc1a58ce53bb0254fb686d5ecfe1"
    },
    {
      "path": ".spt/preserved/307/hertz-w2-regression/negative-289e3cdf-2/observation/receipt.json",
      "description": "Analyzed negative, outer CLEANUP_FAILED; SHA256 b8ba81fadf1eb257715cd8bef5eda4b6c056aa4297229f1ba7c093a6c6e24053"
    },
    {
      "path": ".spt/preserved/307/hertz-w2-regression/negative-289e3cdf-3/observation/receipt.json",
      "description": "Analyzed negative, outer CLEANUP_FAILED; SHA256 6f9c6f0fe7de16142ce91aacc24f9e704221df8bfda823e03f4ad508ce68bf2d"
    },
    {
      "path": ".spt/preserved/307/hertz-w2-regression/negative-289e3cdf-4/observation/receipt.json",
      "description": "Analyzed negative; SHA256 dc9090d36d85c216b1f213cfa146267c05d6ba2febb5bca039b6b7a33bcce08d"
    },
    {
      "path": ".spt/preserved/49-267-refresh-discriminator/windows/discriminator-1/rig/observation/receipt.json",
      "description": "Analyzed negative; SHA256 db66db42de1b21854184c2ac64dafe94abe0250d939baeb9d913f1b3b167a06a"
    },
    {
      "path": ".spt/preserved/49-267-refresh-discriminator/windows/discriminator-2/rig/observation/receipt.json",
      "description": "Analyzed negative; SHA256 1f29b919838bcfe3628e0af7d4aaf6e6dbf267ed83ac8683e7b1675534a60b4f"
    },
    {
      "path": "artifact://1207",
      "description": "Read-only arithmetic output: exact manifest hashes, all 230 ordered post-tag records, completion/send offsets, preserved flags, recovery candidate indices, suffix fractions/runs, poison fields, excluded wrapper schemas."
    }
  ],
  "architecture": "Actual rig receipt has status/config/result/cleanup; result contains checks, refresh.started_mono, baseline, post_refresh, recovery_ms and poison arrays. Wrapper receipts describe guarded commands/processes and must not be replay inputs. Outer status can override predicate_status for cleanup failure. Use retained complete/timely flags; completed_mono minus sent_mono is not necessarily rtt_ms.",
  "report": "## Scope and definitions\nAll exact input paths and byte SHA256 values appear in files. Hashes are receipt-byte hashes, NOT config.sha256 (the executable hash). Recursive receipt enumeration plus broad post_refresh/REGRESSION JSON search covered both requested roots without sampling. All 11 actual receipt hashes differ. Read-only Python -B processes performed JSON arithmetic/hash comparisons only; neither imported nor invoked the rig/analyzer. No source/receipt/lifecycle/pool modifications.\n\nStatistics preserve array order (zero-based tag suffix index). Recovery candidate means first retained timely tag whose completed_mono is within 15000 ms of refresh.started_mono; this is an arithmetic candidate, NOT evaluation of either old/new sustained predicate. Time below is candidate completion offset from refresh start, rounded to integer ms. Inclusive suffix starts at candidate; strict suffix starts at next array entry. All analyzed receipts retain ack_bound_ms=2000 and recovery_bound_ms=15000.\n\n## Exhaustive statistics\n|Case|Retained outer / predicate verdict|Post total/complete/timely|First candidate index / completion offset ms|Candidate-inclusive timely fraction / longest untimely run|Strictly-after candidate fraction / longest run|Replacement / old poison count|\n|---|---|---|---|---|---|---|\n|golden deployah|REGRESSION / result.pass=false|23/23/21|1 / 5281|21/22 (95.4545%) / 1|20/21 (95.2381%) / 1|0 / 0|\n|gate-228 rig-1|PASS|23/23/22|1 / 8422|22/22 (100%) / 0|21/21 (100%) / 0|0 / 0|\n|gate-228 rig-2|PASS|23/23/22|1 / 3188|22/22 (100%) / 0|21/21 (100%) / 0|0 / 0|\n|candidate-1|PASS|23/23/22|1 / 3485|22/22 (100%) / 0|21/21 (100%) / 0|0 / 0|\n|candidate-2|PASS|23/23/22|1 / 8094|22/22 (100%) / 0|21/21 (100%) / 0|0 / 0|\n|negative-289e3cdf-1|CLEANUP_FAILED / PRECONDITION, exit 3|N/A: result absent|N/A|N/A|N/A|N/A|\n|negative-289e3cdf-2|CLEANUP_FAILED / REGRESSION, exit 3|23/0/0|none|N/A|N/A|1 / 0|\n|negative-289e3cdf-3|CLEANUP_FAILED / REGRESSION, exit 3|23/0/0|none|N/A|N/A|1 / 0|\n|negative-289e3cdf-4|REGRESSION, exit 1|23/0/0|none|N/A|N/A|1 / 0|\n|discriminator-1|REGRESSION, exit 1|23/0/0|none|N/A|N/A|1 / 0|\n|discriminator-2|REGRESSION, exit 1|23/0/0|none|N/A|N/A|1 / 0|\n\nFor all five analyzed negatives: full post timely fraction=0/23; longest consecutive untimely run=23; all completion timestamps/RTTs absent; recovery_ms=null. All retain real_refresh, baseline_12_timely_tags, controller_closures_detach_only, reader_healthy=true; ack_recovered_within_bound, sustained_timely_ack_and_state, all_post_tags_accounted_within_bound, zero_replacement_write_poison=false. No after-recovery fraction exists: do not manufacture a 0/0 suffix.\n\nAll four positives have exactly one untimely tag, index 0; every tag index 1..22 complete and timely. Golden has untimely indices 0 and 14; every other tag complete/timely. Full-post fractions: positives 22/23 (95.6522%), golden 21/23 (91.3043%); longest full-post untimely run is 1 in all five. Golden retained recovery_ms=null despite ack_recovered_within_bound=true: do not treat recovery_ms as the absence of a timely candidate. Positive retained recovery_ms matches candidate completion offsets above.\n\n## Recovery/tag measurements\n|Case|index 0 sent/completed offset ms; RTT ms|candidate index 1 sent/completed offset ms; RTT ms|\n|---|---|---|\n|golden|2500/5281; 2765|4078/5281; 1203|\n|rig-1|5453/8422; 2906|6719/8422; 1656|\n|rig-2|625/3188; 2547|1813/3188; 1375|\n|candidate-1|969/3485; 2500|2125/3485; 1360|\n|candidate-2|4453/8094; 2750|5266/8094; 1937|\n\nGolden index 14: sent +17515 ms, completed +19531 ms, RTT 2016 ms. Its strict after-recovery sequence is timely indices 2..13, untimely 14, timely 15..22. Golden indices >1000-ms RTT are 0,1,4,5,6,7,14; in particular index 7=1187 ms, which the RCA's shorter jitter list omits. Complete/timely flags must not be recomputed from completion minus send: candidate-2 index 1 has completion-minus-send=2828 ms but retained RTT=1937 ms and timely=true. Similar observation lag occurs elsewhere.\n\n## Extra controls and schema traps\nnegative-289e3cdf-1 is a SIXTH retained negative attempt beyond the RCA's five analyzed controls. It failed startup with RuntimeError('spawned broker image mismatch'); receipt contains no result/post data and predicate_status=PRECONDITION. It cannot supply a refresh predicate replay; retain and explicitly classify it instead of claiming it was REGRESSION or silently excluding it. negative-2/-3 genuinely contain analyzed regression data despite cleanup-overridden outer status.\n\nThree nested receipt copies exist, not extra independent cases. JSON-object equality against their actual rig receipt was observed true for:\n- .spt/preserved/49-267-refresh-discriminator/windows/candidate-1/boundary-analysis.json:receipt (container SHA256 b294d419778555056ebf3a751d9fe384bf8d6fae32d222b1a18203ce0f65efd4)\n- .spt/preserved/49-267-refresh-discriminator/windows/discriminator-1/boundary-analysis.json:receipt (container SHA256 5d2f9f6a845ff5e375d822dfa258263819e15c01f149a5bc4036a0d00d17c4c3)\n- .spt/preserved/49-267-refresh-discriminator/windows/discriminator-2/boundary-analysis.json:receipt (container SHA256 6602be76df03d5e5735a51b9016e52a8019e0dcee027d6bf02dc618ce7d79a3a)\n\nExcluded guarded-wrapper receipt.json paths under .spt/preserved/49-267-refresh-discriminator/windows/: build-probes, build-probes-2, build-probes-3, build-probes-noack, candidate-1, candidate-2, checks, discriminator-1, discriminator-2, docs, final-seams, pool-claim, pool-release, pool-release-2. Each directory's direct receipt.json records commands/cwd/driver_exit/processes/source snapshots, not result.post_refresh. Also excluded unrelated alternate uppercase receipts .spt/preserved/307/hertz-w1bc/w1b-run1/RECEIPT.json and .spt/preserved/307/hertz-w1bc/w1c-run1/RECEIPT.json: w1b/w1c observation-lane schemas, not the refresh_freeze rig schema. DELIVERY-INDEX.json is a summary of negative-4; job/nextest logs embed already-retained receipts and are not independent observations.\n\nReplay handoff: ten analyzed inputs (golden, four positives, five negatives) plus explicit non-replayable startup-attempt accounting. This inventory makes no claim about a new predicate verdict or tolerance choice; Main owns those."
}