The suggested solution appears to monitor for the presence of the ntpd process, but I'm interested in the daemon's stratum and difference from the monitoring system's time - numbers provided (among others) by the ntpdc "peers" command.
I visualize the ntpd servers as blocks of wood floating off a dock, and tied to each other by threads. The dock is stratum 1, blocks tied to the dock are stratum 2, blocks tied to stratum 2 blocks are stratum 3, and so on. Occasionally a wave will come along (network disruption or pervasive clock interrupt misses) and snap the threads connecting a set a blocks, and they'll go drifting off on their own. I want to be informed when that happens. I want to know when NTP daemons change their stratum, and I want to know how far off the monitors time each daemon is.
With nagios, I ran an ntpdc command on the nagios server and lightly massaged its output for consumption. It doesn't feel right to ssh into a Linux box for each data point; is there something more appropriate?