If we assume you're using ntpd on your instances, and that they're all configured to peer with the Google Public NTP service, then you could set up some monitoring that looked at the ntp peer stats regularly for each of your instances.
ntpq -c pe from the commandline will report the performance of the peers on each instance.
ntpq -c sysinfo will show the local instance's performance.
A script that extracted these values and put them into something like Prometheus could be done in Python. I'd be surprised if there wasn't already such a thing on the internet.
But if you want a quick answer, running ntpq -c sysinfo on all your instances with a command like clusterssh or mcollective would get you there, but relying on regular manual work to maintain it.