Although we don’t know for sure how much data Sense is sending to the back end systems, I doubt it’s going to be sufficient to tax even a slow urban connection unless that connection is also being used to push lots of data into the cloud at the same time. It would probably struggle with a rural DSL at the limit of copper connectivity (say 1 mb/s upstream)
We do know that Sense reboots after a longish period as part of its strategy to restore the connection, perfectly reasonable, but appearing to dump buffered data in the process.
I’ve seen several outages on my data, due to my WAP location, now fixed, that also aren’t explained by a reboot strategy, so I do think there is some work to do on handling the buffer.
I’d suggest that the engineering team look at “warm” restart strategies if not already in work. These can preserve data and reinitialize external connectivity, so that the router reassigns a new IP address for instance, of a conflict was causing the loss of connectivity.