Well, diffusion in air is orders of magnitude faster than diffusion in water so I'm not sure that would give you much advantage over the current designs.

The mass transfer across a gas-liquid boundary and within the liquid phase itself are probably the rate limiting steps.

Reducing the volume of liquid and increasing the mass transfer across the gas-liquid boundary would probably have much more impact. The mass transfer across the boundary can be enhanced by using the proper membrane or by maximizing the gas-liquid interfacial area. In other words, rather than shrinking the volume of gas, you’d be better off with a very wide, thin disk of reference solution in contact with the gas phase or a semi-permeable membrane.

I think I saw some comments along this line of thought by Hoppy or someone in another thread where they were looking at using a membrane rather than an air gap between the tank water and reference solution.