I think there was something hokey about the power connection. It sits on top of an APC UPS in between outlet rows and is powered by the battery side. Since I have other things also plugged into the UPS, I goof with them on occasion. I've made the modem reboot by leaning it to the side to unplug/replug a printer and the cooling fan for the modem and also swapping LAN cables. After pulling and reseating the power supply plug, it seems to be doing better.
I've also moved my router very close to the modem with a short cable rather than having it up high and 6' away with a longer cable so maybe that had something to do with it too.
Third thing I found is my coax was a quarter turn loose. It's probably been that way since the coax/grounding block was changed out last year which is what eventually solved the persistent SQF 15 error I was experiencing for so long. In fact, I don't think I've ever had to disconnect coax and/or power since that was done. Doing the cable dance by disconnecting both used to be the only way to recover from the SQF 15 error.
I know for a fact everything is grounded properly. When I got my original Gen4 install, I had a real pi***** match with the installation company who (twice) refused to ground it to Hughes specs. Of course, I wasn't allowed by Hughes to do it myself so this required a completely different service company to come out and complete the work.
A couple of thoughts on your modem placement and power connection.
How good is the airflow to the modem ? They generate some heat and that needs to be carried away. Is it in a place where you can see the modems front panel LED's ?
If a overtemp occurs the power LED will turn red prior to doing a shutdown.
As to being on the battery side of the UPS ...
That is likely to isolate the modem grounding wise. The ODU and Modem need to be at same grounding point.
Also, UPS's (except for very expensive models) put out a modified sine wave.
It is possible that the modem isn't liking the "flavor" of its input voltage both in waveform and perhaps current at a given instant.
The speeds held up reasonably well for a Saturday but started crashing after 6. I set up another auto testing routine beginning at 7 PM to run on 15 minute intervals throughout the primetime period. I set it to allow testmy to choose it's own file size and wired direct from my main computer to the modem.
As usual, the first test shown in the series is the last test shown in the previous series.
Will run an hourly test overnight again.