With the adjustable delay line box installed in the 55MHz modulation path, I've measured the PRMI sensing matrix as a function of delay / relative phase between the 11MHz and 55MHz modulations. The relative frequency difference of 44MHz tells us that this should be cyclical after ~23nsec of delay, but losses in the delay cable change this; see Koji's elogs about the modulation cancellation setup for details.
TL;DR: Nothing really changes, other than REFL33 optical gain. MICH/PRCL angles remain degenerate.
The results aren't so surprising. The demod angles for the 55MHz diodes don't even change, since the same 55MHz signal is used for the modulator and demodulators, so delaying it before the split should go unnoticed. Most of these measurements were made during the same lock stretch, PRCL on REFL11 I and MICH on AS55Q.
The only signals we would expect to change much are ones that have significant contriubtions from field products influenced by both modulations. None of the 1F PDs are like this, nor is REFL165. REFL33 is the odd man out, where the +44MHz field produced as a -11MHz sideband on the +55MHz sideband beats with the +11MHz sideband (and the same with the signs flipped). I made a simulation for the 40m poster at the March 2015 LVC meeting, but I don't think it ever made it to the ELOG.
Here are the results for the 0ns and 4ns cases, as an illustration of what changes (REFL33), and what doesn't (everything else). Again, these are calibrated to Volts out of the analog demod boards per meter of DoF motion.
So, since REFL33 is the only one really changing, let's just look at it by itself:
Qualitatively, the change in magnitude looks similar to the simulation result. The demod angles fall by some roughly linear amount. The angle difference is even more stationary than predicted there, though.