eng-0096
window recorded
citable URL: https://halobench.com/records/eng-0096/ — this address never moves; the anchor /records/#eng-0096 keeps resolving
22.72 Whmean 172.6 W − 10.1 W idle floor (box-idle-all-empty) = 162.5 W attributable
window 2026-08-15T13:47:40Z → 2026-08-15T13:55:34Z · provenance recorded
method wall-meter · aihydra energy sensor via Home Assistant (counter-difference on sensor.hardware_ai_hydra_energy) · no contention
covers run-0286
EOS-cliff cause-isolation matrix, avoidance probe 4 / server restarted every 3 requests on the failing baseline config (run-0286, cfg-0075, 15 requests across 5 fresh server processes, 1 degenerate — down sharply from 11/15 continuous); wh_per_task/pence_per_correct_answer left null (guard run). This window includes 5 model-load overheads (~9-13s each observed in the isolation log) inside the wall-meter total, so it is not directly comparable per-request to the continuous-session cells (run-0277, run-0283) despite similar total Wh — reload overhead is part of what restart-as-mitigation costs operationally. Counter samples: start 19.5998108834028176 kWh (shared boundary with eng-0095's end), end 19.6225305050611476 kWh (14:55:39 local, 5s after window close — nearest recorded tick). Retroactive join computed 2026-08-15.
Wh, never joules — read as a cumulative-counter difference across the window; delta over the idle baseline is the only figure that means anything