On a final note, we crawled our client systems checking for misbehaving 
qpidd, Firefox and Thunderbird instances and found a 73% incidence of 
the problem. That rate can be influenced, of course, by the fact that 
not all stations had said software running at the time and some have 
been rebooted since the leap second, but this is significant. We've 
confirmed, however, that not all systems were affected, though -- but 
that the problem is more widespread than we'd initially thought.

No crashes have occurred, but telling all systems to:

date; date $(date +"%m%d%H%M%C%y.%S"); date

fixes this across all three applications immediately, without a reboot 
and with no negative impact. On systems with 6 or more processors this 
problem was difficult to diagnose initially, just because the percentage 
of waste appears to be non-problematic when simply checking total system 
resource usage instead of looking at per-processor statistics (the waste 
on a 12-core system will average around only 7 to 8%, which seems like 
nothing, until you consider what that really represents).