Actually, this has been done before, by Claudio i2NDT (check it out!).
So, what's the big fuzz about another approach? Well, testing with transmitters would not allow for testing active antennae, obviously. So, what will it be, 2 receivers of the same make? Well, maybe. Tolerances could still play tricks and mess up the measurements.
If you think of it, the solution is as obvious as it is simple... almost a miracle that nobody was writing about it before: use an I/Q-SDR direct conversion receiver with some slight modifications.
Such a receiver actually consists of two identical receiver sharing a local oscillator. The 90 degrees phase shift
Even the cheapest of I/Q-SDR receivers aims for the I and the Q channels being matched as good as possible. Usually, low tolerance parts are used all over.
As modification I would consider the following (once again one of my lists):
- add a switch to toggle between the usual 3dB splitter and two individual (crystal) filtered front ends
- configure spectrum-lab to show left and write audio channels in individual waterfalls