ip_conntrack: table full

hi,
we've a fully updated rh8.0 firewall with kernel-2.4.20-18.8,
iptables-1.2.6a-2. we got the following error about once a week:
-----------------------------------------
Jun 13 05:21:41 portal kernel: ip_conntrack: table full, dropping packet.
Jun 13 05:21:47 portal last message repeated 10 times
Jun 13 05:21:51 portal kernel: NET: 6 messages suppressed.
Jun 13 05:21:51 portal kernel: ip_conntrack: table full, dropping packet.
Jun 13 05:21:57 portal kernel: NET: 3 messages suppressed.
Jun 13 05:21:57 portal kernel: ip_conntrack: table full, dropping packet.
-----------------------------------------
in this case we cant do anything just reboot the server. the strange is 
that in this case:
wc -l /proc/net/ip_conntrack
is about 3-400, while during the normal operation it's about 1000-1500. 
so I don't know what can be the problem in this case. what can I do to 
find out the reason for this problem? and how to solve it?
thank you for your help in advance.

-- 
   Levente                               "Si vis pacem para bellum!"