Based on my experiences, the missing time slices in the history timeline were treated as zeros for the purposes of the Always On calculation. Throughout 2017-2018, big downward glitches in exported hourly Always On were a strong predictor that a history timeline had a dropout.
While it would seem like a no-brainer to exclude the “missing” data points from the Always On calculation, I think there is something trickier here at play. Based on what I have seen, many of the “dropouts” in the Total Usage (not smart plug) history timeline were not really dropouts, they were errors/glitches in Sense’s realtime conversion/data management from the raw data, to the history timeline. Even though the history timeline looks like it is simply displaying data from the monitor, that isn’t the case - the raw data (microsecond level measurements) goes through a number of calculations before it becomes the 1/2 second data updates we see in the Power Meter.
I say this because I have worked with the Sense support guys to try to reduce the number of dropouts I was experiencing in late 2018. Two themes emerged:
-
At least for me, the dropouts mostly correlated with opening up or running the Power Meter running live on one ore more devices. Turn off the Power Meter, and refrain from opening it as frequently as I was, back in mid-late 2018, and I would see fewer dropouts.
-
During a couple of deep dives by the support guys to sort out the source of these dropouts, they were a bit confounded. It seemed that my raw data at the Sense mothership was intact, but portions of the history timeline had been zeroed out. In my mind, likely related to memory management of near-realtime display data in the Power Meter vs. long term storage of the history timeline.
I still have regular shorter dropouts, but they are typically short enough to not affect my always on.