[FW-1] db2 errors / too many concurrent connections? (4.1. sp5, ipso 3.4.1)

[Security Guy <firewall_security@FW1-NOSPAMHOTMAIL.COM>]

Here's a problem that has been steadily growing worse for several months:

 

                          {Internet}
                                |
[webserver 1]  --         |
[webserver 2]   ------| DMZ |
[webserver 3]  --         |
(2 nics>                    |
 1 ext, 1 to dmz)        |
     |                       
                  {Internal Network} 
                            |
      Mainframe           Unix Box 
     (db2 connect)      (for images)  

Connectivity between our DMZ and the internal interface periodical drops, approximately once a week. The only way to restore connectively is to re-publish the rules and reboot the offending web server. We don't believe the problem is with the web servers, we've added another & the problems remains the same. Is this an issue with the table.def file filling up? Correct me if I am wrong, but can checkpoint FW1 (4.1 sp5) handle more that 25,000 concurrent connections?

I'm guessing the connections table (table.def) fills up and the only to purge the connection is to flush the table. We currently have a ticket open with Nokia, they are building a script that will clear the table. Is going down the right path? Another option is to segment this troubled portion of our network with another F/W but that will require a significant amount of work.

Any help is appreciated.

Thanks