Ok. I've spent a lot of time tonight playing with this idea.
I made three videos.
 
Basically, if I were to use this at all, I would not invert the audio where I'm listening to the audio being removed. That's just weird and not very useful imho. It's not like a De-Esser where you want to hear the sibilance being trimmed. It's a compressor.
 
I might however use it to accent the compression so I could dial in the settings.
 
Let's say I get my compressor where I think it sounds good and it's dipping down at worst to -7dB. So I introduce the phase flipped track 14dB below that level and now, I can clearly hear what the compression is doing by making the track 'shallower' if you will. By making the track shallow, I accentuate the actions of the compressor and I can then maybe tweak the attack-release-ratio-threshold to dial it in just so.
 
But inverting the tracks so that the compressed track is the softer of the two tracks. I'm not sure that gives you anything valuable.
 
And dialing the gain so that the two tracks are close is even more useless because then you aren't even sure which track is louder. You get all kinds of garbage noise, but you don't know what it means at all.