Would there be and detrimental effects to a bipolar suppled amplifier which sees two different supply output impedances on the negative and positive rails (i.e. whatever NYD is powering with this circuit)? I could see this happening as the two transistors are basically changing supply ouput impedance to correct voltage differences. Someone mentioned oscilations when driving capacitave loads, but I could see adding a few large storage caps after this stage being benifital to lowering transient output impedance of this stage.
There seems to be a lot of class A power being wasted in this stage. Of course a tightly regulated switcher would be benifitial in this case as you can achieve control loop bandwidths well above the audible frequency range (especially current mode). Of course now, you have to deal with switching noise, but your efficiency would be much higher. You could implement a solution with two control loops (one for +, one for -) but this would be much higher parts count, and a probably significant amount of design time. With swithcing noise in the well above 100kHz range, i dont see how this could negatively effect your fidelity. Of course, I know we like to keep our supplies linear around here for the most part. Call me crazy, but I'm all about the efficiency.
-Mike