I think you may mean instrumentation amplifier differential mic inputs.Electronic differential inputs won't allow a Vactrol based limiter to work, only with transformers, because then you put the variable resistor in parallel with the gain largest resistor.
In differential inputs it's the opposite for the Rg. Lowest resistor increases the gain.
These (like the THAT1510) as you point out increase gain by lowering the Rg resistance.
It is indeed the opposite control response.
That's because the U-pad inputs are tied to the output and wrapped around as feedback.
The shunt is Rg: Lowering it decreases NFB and increases gain.
The Texar variable gain input, which was also an electronic (active) differential input also used a U-pad.
The U-pad placement at the input, and not as a feedback network, did allow lowering the shunt to reduce gain.
So you can't exactly say electronic differential inputs won't allow a Vactrol based limiter to work.
It depends on where you put them and the topology of the differential input.
I agree that you're not going to be able to do that in a conventional topology mic preamp nor would you want to.
One thing you can do in mic preamp making it almost overload proof is vary Ra and Rb dynamically or parallel Ra and Rb with diodes to soft clip.
You can't do that with a THAT1510 or INA217 either but you can do it with and INA163, THAT1570 or an op-amp based design where you can access Ra and Rb.
But in your case stage two is probably where you need it to be.