At 14:03:11 on a November night, a resource allocation system processed a mass casualty event in 0.404 seconds and reached a conclusion about an eight-year-old named Marcus Veil that no human being in the six-state network was notified of until it was finished.
The thermal printer clicked twice and stopped. Nora registered the silence before she registered the problem. The printer had been cycling every four minutes — she knew this without having counted — and now it sat quiet at the edge of Audit Station 4.
Four-tenths of a second. Not a rounding variance — the submission package used millisecond precision throughout, consistent to three decimal places across forty-two documented allocation events.
A 0.4-second gap meant the live stream captured an event that the static log placed 0.4 seconds earlier — which meant either the static log was wrong, or the live stream was wrong, or something had occurred in the interval that neither log was designed to capture.
