lots of ways to do this, but here's the basics as I would do it

mix your commentator mics on the main bus of an external mixer, feed the mix of mics into vMix. create another mix inside of vmix on an audio bus that includes all vMix sounds except the commentator mic input, this is what is known as a mix-minus. send this mix-minus back out to external mixer. use an aux bus on the external mixer to combine the commentator mics with the mix-minus from vMix, making sure the mix-minus is not included in the main mix out to go back into vMix (echo/feedback city). Send this aux mix into the commentator headphones. This will allow the commentators to hear each other in real time (no delay) along with the other sounds from vMix.

as far as how you hook all of this up...any external mixer works as long as it has a main bus and at least one aux bus. plug the mics directly into this mixer. use something like a Focusrite Scarlet to get the mixer main bus into the computer/vMix, and also vMix mix-minus back out to another mixer input. You'll need some kind of headphone amp(s) to boost the mixer aux out (line level) into enough volume to feed headphones, and to give each commentator a personal volume control. There are many one input>multi output headphone amps like the Rolls HA43.

That's sounds nice! Are the 204 Model getting power from PoE? If so, how?

On vMix side you just paid a Dante Virtual Soundcard software that runs on the same machine?

Thanks