There are multiple places where a timeout can happen.
If the users are dialing into a concentrator (such as a Shiva box), there is a timeout parameter in their account information there. The same applies if they are dialing into a Microsoft RAS server, but the timeout is global.
If the users are dialing into an ISP and setting up a VPN to your network, there are timeouts at their dial-up adaptor, at the ISP, on their dial-up entry for the VPN, and on the VPN server.
Some possible solutions:
1. Run the jobs with |at now rather than in the background. This may or may not work depending on whether or not the jobs really need interaction. (You will have to enable the users to use the 'at' command.)
2. If you have a Terminal Server available, have the user run a Terminal Server session (set the user to suspend on loss of signal). From the session, telnet to the unix box and run the SAS job. If their connection dies, it does not matter, since the SAS job is being run by the Terminal Server. They can dial back into the Terminal Server and resume the session.
3. Put a PC per remote user in the shop. Have the user dial into the PC with some remote control software (PC Anywhere or similar), and telnet or X into the unix server from there. Make sure the remote control software does not insist on a logoff or reboot at the end of a session.
Good luck.