Generally speaking the routing table on the Windows Server decides where Internet-routed packets go. So the solution must be sought there.
In OpenVPN you can try to specify a particular internet gateway as the gateway for specific subnets, but that will fail if the OS routing table decides that to reach it, it must go through a particular internet connection.
On Linux you can do a routing policy whereby traffic coming from a specific interface can be configured to go through a specific interface. So you can lock the OpenVPN interface traffic to be routed out through one particular Internet connection.
If the internet gateway isn't directly on this Windows Server but is instead on a separate router that has multiple WAN connections, the solution must be sought in that router. To try and configure something like perhaps a VLAN with a specific internet gateway, or policy routing like I described.
This is going to be very difficult to resolve within the VPN solution itself.
