The creature active overnight
Nineteen hours awake, and the three thresholds that made sleep impossible.
The first overnight run on v06.5 produced a strange result: the field never slept. Not once, across about 19 hours. The sleep log had zero entries.
Traced it to the three sleep gates. The quiet threshold was 0.055, left over from before the predictive cell model. But the weather anchor holds a steady pressure at all times, and the overnight median top-cell pressure was around 0.6 with everything else reading zero. The condition could never be true. The other two gates sat flat all night, equally unreachable. All three doors to sleep were painted on.
The fix: measure what calm actually looks like now, and set the threshold just above it. Quiet hours peaked around 0.68, real activity started near 0.9, so the threshold moved to 0.7. The creature sleeps now.