I make use of pfSense in many of my network designs. It’s lightweight and versatile, and runs well as a VM under low-to-medium usage.
I often connect networks together using a pfSense VM and OpenVPN, and this design usually involves NAT. As a result, I end up with a NAT rule somewhere with a source address of “OpenVPN”, as illustrated below.
As I’ve discovered to my pain, this is a “Bad Idea(tm)” if you have multiple OpenVPN instances. pfSense will alternate the source NAT between your various instances, making at least 50% of your sessions through the NAT spectacularly fail.
So how to do NAT over OpenVPN interface then, if you can’t select a specific interface? Avoid using the interface dropdown, and instead explictily state the NAT address as a /32 subnet in full CIDR notation, as illustrated below.
Provided your rule only matches the traffic you want it to match (i.e., you’ve specified a source and destination in the NAT rule), your OpenVPN NAT will now perform the way you’d naturally expect it to.