In my environment, when we are at home connecting to our intranet using VPN, we are getting two IP addresses registered with DNS. One is for the WiFi hardware port (hidden behind the VPN), and the second is for the VPN client (the good one). You can see this by running the nslookup hostName command on a login node of the cluster. When trying to connect back to the client side (Mac, Windows, Linux), it is random if DNS will provide the incorrect IP address to ParaView for the reverse connection port connect.
We found the solution today. ssh into your cluster login nodes, and type a printenv command. You will find an environment variable called SSH_CLIENT. This will include the ip address that is making the ssh call (i.e., the client computer). This can then be included in the scrips that start the server connection.
Alan