I can't speak to the accuracy of the diagnosis, but the claims about NTP are bizarre, and to the best of my knowledge, wrong. There's nothing specific about the times the incidents cluster around that would have anything to do with NTP. It doesn't work like that.
>The precision of outages (at :29 and :44) matches a network-synchronized clock (NTP).
I think this just correctly points out that if the trigger was something unsynchronized like animals chewing on wires or someone digging underground, you wouldn't have 61% of events occurring at these two second markers. Even if the trigger was something digital but on a machine that isn't NTP synchronized, you would eventually have enough clock drift to move the events to other seconds. 61% combined at two markers (exactly 15 seconds apart) strongly suggests synchronized time.
