NTP synchronization problem

Dear Managers,

We're running NTP services on a 4-node cluster (Tru64 v5.1B w/ PK5).
For some reason, after a few weeks we noticed that each node has a
different system time (between 10 seconds and 20 minutes difference from
the time server).  Each node has the same ntp configuration
(/etc/ntp.conf).  

We can force a client-server time synchronization by restarting the
xntpd daemons on each node and, that works fine for a while
(/usr/sbin/xntpd -g -x -c /etc/ntp.conf).  Supposedly this
synchronization should happen automatically every so often.  Also, time
should be corrected little by little automatically and not the opposite.
Is there a way to 'force' this time synchronization after X amount of
time?  Maybe by adding some option(s) to the /etc/ntp.conf file or
starting the xntpd daemons and using certain parameters???

Regards,
Orlando