[Cieplak]

I've had weird experiences with long-lived ssh tunnels and have often resorted to initializing the ssh tunnel from a python script and using a heartbeat to manage the state of the tunnel to terminate and reinitiate the tunnel process if it hangs for more than 10 seconds. That only works if the applications working over the tunnel can tolerate dropped TCP connections, eg HTTP + some connection retry logic which opens a TCP socket per request vs rsyslog (without RELP) which trusts TCP's ack mechanism.

One hypothesis for your firefox experience:

http://askubuntu.com/questions/344863/ssh-new-connection-beg...