It is possible. It's a two step process:
- Configure the RSCD agents in each network so that they communicate over the specific port required
- Configure the secure file on each of your central application servers so that they know which targets communicate over which ports.
You might also have to configure routing rules to these hosts, depending on whether or not your central application servers can resolve the targets in the remote network segments.
Either way the easiest solution in this situation is to have a standard RSCD agent port across the environment. If it can't be 4750, then it would be better standardizing on something different instead of using and maintaining communication across various ports.