I've had a few instances of "instability" with timesync.

Have a NW 6.0 SP5 Single Time Server getting its time
from an NTP server maintained for the campus network. Four
Netware servers look to the Single server as their time source.

Occasionally, the Single Time Server seems to be caught bouncing
back and forth unable to synchronize to the NTP Server.

ie, the TimeSync Debug Screen will show continuous polling,
with one poll reporting an adjustment requested of, say,
2500 milliseconds (or more) and the next one an adjustment
of something like -2500 milliseconds. This may go on all night
(and result in 500 Health Messages to me).

It does not ever seem to recover on its own. One time I increased
the Synchronization Radius setting by 1 second and it stopped.
Another time increasing the Polling Count to 10 seemed to fix it.
A third time, Setting the Restart Flag once was enough.

Anyone seen something like this? Any suggestions on settings, etc,
to avoid it?