Wow, I never thought about DC "biasing" a capacitor before. I looked up and found an original document from JBL on it and there seems to be at least a little bit of horseshit/random audio voodoo mixed in there but I guess could potentially work.
http://forums.klipsch.com/forums/storage/3/860249/page10.jpg
http://forums.klipsch.com/forums/storage/3/860250/page11.jpg
http://forums.klipsch.com/forums/storage/3/860252/page12.jpg
He's comparing it to biasing transistors into Class A but it doesn't seem like a good analogy, since the transistor is an active device that needs to switch on if it isn't biased into class A whereas a cap is, well, passive, and "biasing" pretty much just means charging it all the way up with DC unless I'm missing something here. The piezoelectric effect seems to me like random horseshit/voodoo unless you're using ferroelectric dielectric caps, and even then I don't know how much the capacitor would actually "move". Who knows though, I haven't done the measurements, but neither has he.
Needing double the capacitance and double the caps (which means double the ESR?) seems like a bad trade-off for supposedly getting rid of whatever effects dielectric absorption has, though I think I saw some other design with just a 2.2M resistor? Have you thought about implementing a Zobel network or altering phase?