As I was curious how the leap-second of was handled, I dug into the manuals of NTP. To monitor all of this, I built (mostly copied from here) a script so that Cacti could create a graph to monitor offset, jitter and stability for NTP.

The leap second generated this log-message:

   Jan 1 00:59:59 server kernel: Clock: inserting leap second 23:59:60 UTC

This is the graph for the leap second:
NTP statistics graph

It shows nothing special, the leap second occured at 01:00 as this is GMT+1. It's created with Cacti and this script (rename tp ntp.pl).

The original script had a few warnings and some minor errors and I've rewritten it to match the output of ntpq. Some output is multiplied by 1000, so when you graph this you should divide it by 1000 to get normal values.

This is the graph for a month, it shows a spike in the stability almost exactly once per week. I'm not sure where the spike comes from, it's not exactly every week at the same time.
NTP statistics graph

The average offset is 350 milliseconds. The spikes are caused by an increase in cpu-usage at that moment, these spikes are caused by the weekly scan for virusses.