do you encounter any issues with the Default Settings?
Basically I recommend to leave the default settings in place. They are based on our experiences in the past and should pretty much reflect what we expect to be the "most optimized" settings for MOST customers (of course the can not be suitable for everybody).
We seem to get timeout issues when users are accessing services over the internet particularly when running large queries or reports on back end servers and the background activity takes a long time before presenting the information either via a java client or .csv report. We tend to get connection lost errors. The technical team are leaning towards tunnelling timeout setting on https/SSL sessions.
There is a setting to allow Web Gateway to send keepalives...I'm not at work just now but if you search for it that might help.
Incidentally I have just cleared up a problem we had where one site we use was timing out.
The host company said it was a "bandwidth" issue. When i pointed out I got the same errors from home, they decided to look closer.
Result ? They applied some code fixes and the site now works.
If possible try to test the site outside your infrastructure.
I have seen this before, especially with reports being generated on a server and the server not coming back until the report is completed. Basically it is the servers task to send some bytes to the client to keep the connection alive, because otherwise the connection will be dropped after a while. However Webwasher also has some timouts which drop idle connections. I think especially about the "Dead Client Timeout". Have you tried to increase this? You may double it and then give it another try. If you see it takes longer until the timeouts occur you are on the right track.
If the server does not send us any indication that it is still working we have no way to keep the conection opened.