Mike - indeed
obviously you have some knowledge in this area
Back in early late nighties when I worked on early FPGA based modulator designs I still recall how when we changed from a second order to third order modulator the third order sounded worst then the 2nd order despite its better simulated results. The 3rd order lost the "Dynamic" realism of the simpler 2nd order Mod. much in the same way I have a problem with PCM verses DSD.
ESS has a decent video lecture on the net where they discuss the care they have taken with there modulator design to "Constrain" the State variables during transient changes - very interesting as this closely follows what we discovered.
One of our solution is to use cascaded 1st order modulators (AKA a MASH type structure) where each modulator loop is therefore inherently "Bounded" between changes of state....
Also we would like to try "predictive" modulators...