Virtualization: HOW can I connect a virtual NAT through a VPN interface?

OS: Debian Bullseye

VPN: OpenVPN 2.5 - (redirect-gateway def1)

Virtualization: KVM - qemu/libvirt latest version

Using Whonix for KVM

Issue:

When starting qemu's default network, it automatically implements iptable rules allowing all libvirt on OUTPUT, FORWARD, INPUT chain anywhere on eno1 and FWD 192.168.122

Code: -A LIBVIRT_FWI -d 192.168.122.0/24 -o virbr0 -m conntrack --ctstate RELATED,ESTABLISHED -j ACCEPT

-A LIBVIRT_FWI -o virbr0 -j REJECT --reject-with icmp-port-unreachable

-A LIBVIRT_FWO -s 192.168.122.0/24 -i virbr0 -j ACCEPT

-A LIBVIRT_FWO -i virbr0 -j REJECT --reject-with icmp-port-unreachable

OUTPUT Chain

LIBVIRT_OUT all -- anywhere anywhere



FORWARD Chain

LIBVIRT_FWX all -- anywhere anywhere

LIBVIRT_FWI all -- anywhere anywhere

LIBVIRT_FWO all -- anywhere anywhere



INPUT Chain

LIBVIRT_INP all -- anywhere anywhere
This means that in case the VPN disconnects, Whonix traffic would continue over eno1.

It forces me to use VPN-Firewall or similar solutions, to achieve Fail-Closed.

Desired Behavior:

I would like the Gateway to specifically NAT through tun0 instead of eno1, so that a simple ufw rule only allowing out on tun0 would be enough to cut off all traffic in case the VPN drops.

This works *out of the box* with Virtualbox, but I really want to use KVM, if it can be done somehow.

My failed attempts:

Including the following in the external network xml
Code:
<interface dev='tun0'/>

<forward dev='tun0>
Changing iptable rules to drop all packets between eno1/virbrX in FORWARD chain and only allowing tun0.
This simply breaks the entire network without the ability to ping from either the Gateway or the Workstation.

Any ideas?