RID Sensor

Why this build: it adds a record of what the sensor asked its WiFi chip to do just before a reset, and stops a sensor being handed an out-of-date update file. The sensors still reset roughly once every twenty hours each. They come back on their own in about twenty seconds, but the cause is not known. Five of those resets have now been recorded in detail, and every one looks the same: both of the chip's processors stop at the same instant while the WiFi chip's own software is running. Twice, two sensors reset in the same minute, which points at something in the air rather than a fault in one unit. What the record could not say is whether the sensor had just asked the WiFi chip for something. A sensor on this build remembers its last few such requests, each with the time it was made, and keeps them through the reset. The second change: yesterday one sensor was handed a cached copy of its old update file and sat on the previous build for five and a half hours. Each check now asks in a way that cannot be answered from a cache. It went to one sensor first, then to the rest one at a time, and each had to stay up and keep reporting for twenty-two minutes before the next one started.