In any medium to major comms infrastructure project, we'll install as part of the build a NTP clock. Or in English, a black box with a GPS antenna that uses the atomic clock from satellites as the time source for all the various network connected doodads.
Prisons, airports, universities would all have as a minimum at least a master and physically separated failover NTP box solution, as THE master clock is an obviously critical piece of equipment.
The chipsets of choice for all the NTP server's was Trimble. For alarm purposes, we wrote code to poll the chipset for hello, synch drift, synch loss all that sort of thing, and in the process we found the Trimble chipsets would reboot every 24 hours.
????? Trimble supply 10's of millions of their chipsets to people like Garmin, US military, they are a quality chipset. We took lots of tests results and flicked them over to Trimble,
Imagine my surprise, when they came back and admitted quite openly that it was old code that was designed to reboot the chipset every 24 hours to mitigate drift! We asked if it could be removed from the code, as it was generating alarms every night when it rebooted. In facilities like prisons, that's not good for a critical alarm going off in the middle of the night.
Trimble did it all no problem, they are really good to work with. But I still cannot believe squillions of these same chipsets are out there, rebooting every 24 hours still, years after the reliability of sw improved where it is no longer flaky to at least not have pull the plug on it every day!
So, if you have a Trimble GPS chipset that you haven't firmware upgraded since 2006 (for that read nearly all of us) , you'll notice in debug error logs at least once every 24 hours a lost satellite comms link!!!

Behind the scenes of some black magic... still the good ol' reboot.