...
No clear pattern. It's even difficult to see if the nagios ttstation failures line up with the reboots. Past incidences did seem to be at regulat 10-12 minute intervals, but not this time.
The SPOL conjecture
Steve and Chris measured the batteries, looked for problems on the power interface card, and we scratched our heads for a while, then came the real kicker. Steve realized that SPOL had been running while the reboots were happening, and now that SPOL had stopped, the DSM had not rebooted. Since we could find no indication of a problem in the hardware, the current running theory is that SPOL radiation was somehow interrupting power. That also happens to explain unexpected outages in the iss4station Ubiquiti radio. Nothing else at that site has gone down, except for the ubiquiti, with its antenna pointed almost directly at SPOL.
...
No DSM reboots in the following 24 hours, and no iss4station outages, but probably SPOL has not been running.
Mitigation?
If the problem does turn out to be SPOL, and assuming the DSM can be shielded, I'm not sure how we can mitigate the interference with the Ubiquiti radios. The other radios have not had problems, so apparently it's enough to point the radios away from SPOL. However, that would mean introducing another radio which can point at pedestal without pointing towards SPOL, then pointing iss4station and ttstation at that intermediate radio.