Very strange that should be happening, if the call is established then it should not disconnect unless it is forced to as the signalling is complete. I would be very interested to see debug voip ccapi of the call dropping and see what is happening.
Nonetheless, if your gateways or Gatekeeper have IOS above 12.2.2T then you should upgrade to pix 6.3 or above as that supports h323 v 3 and 4 whereas 6.1.4 supports h323 version 2 only.
Although that should not have any bearing on the disconnect of the call as i meantioned. Very strange scenario!
I have almost the same problem: multisite IP telephony centralized call processing through a PIX but no GK. Calls at the branch site disconnects after some 8 minutes (8 min 28 sec axactly). I opened a case (# D878688 ), I collected a lot of debug, sniffer trace etc. no result. I replaced the whole gateway (c3725 IOS 12.2.3), no changes. I also waiting for a miracle...
I have seen your problem before and have come up with a solution that I had to use. The problem is with the H323 timeout setting, unfortunatly the timeout that you adjust before 6.2 does not work. If you upgrade to 6.3 it should take care of it. I had to get a custome baked IOS from cisco to fix it but the 6.3 code is supposed to have it fixed in it. For you notes Cisco sent me PIX code 6.2(2).
Are you getting this error “Installer User Interface Mode Not Supported. The installer cannot run in this UI mode. To specify the interface mode, use the -i command-line option, followed by the UI mode identifier. The value UI mode identifiers...
The below trick might come handy when you have to add a new node to a cluster but you don't have or is unsure of the security password for the publisher. This procedure has been around for ages.
1) Login into the CLI of the Publisher.