Yes, the requirements for the input/buffer stage are actually quite modest, not even a headphone amp, really. But the current fad for floating 12 or 15V supplies from a switching wall wart limits the output swing and current available. Barely enough for an op-amp (+/- 6 volts), plus losses from local sub-regulation.
It makes sense for the op-amp (or discrete circuit) to be fully isolated from the Class D switching module. The Class D module generates program-modulated switch noise ... it's effectively a low-frequency AM transmitter contained within the chassis. That's where the efficiency comes from, after all ... when there's no program material, switching is still going on at 200~500 kHz, but no power is going through the output devices, and very little is drawn from the support circuits. There's no residual Class A idling as there is with Class AB amplifiers. The output devices are either on or off, with only extremely brief switch transitions.
As program material level increases, more power and switch noise is created by the output switcher, and filtration demands on the speaker output and AC power supply increase. It is not trivial to silence a 200-watt AM transmitter in a can ... that energy is going to escape any way it can. Through the speaker wires (which make a great antenna), through the AC power cord, and even through the input jacks if it can find a way. Or leaks in the metal can itself. The adjacent linear audio equipment will have varying levels of tolerance for nearby RF emitters, which not usually tested in most test scenarios.
@lynn_olson Just to set the record straight, I don't think any class D designer or those planning to use and existing class D module would ever consider using a wall wart as a power supply for the input buffer!
The switching noise is far lower than you suggest! In fact so low that many tube amps radiate more noise (due to their rectifiers). As a result, its quite practical to put the input buffer opamps on the same board with the class D section as a complete module and still have it so quiet that you'd struggle to hear its noise floor on a horn system.
You really do want the switching noise quite low because if it radiates it can mess with digital devices. Noise is really a matter of good layout. Typically its nice to have the radiated noise about 60dB below the level required to meet EU standards to obtain the CE mark.
We used an input transformer on our prototypes. It worked quite well. But they are impractical due to reflected impedances- how well they drive the load has a lot to do with the output impedance of the source driving the transformer. Since the preamp is what we're talking about, the result would be variable; in some cases the transformer would drive the input of the comparator circuit quite well and in other cases, not so much, just because of the preamp driving the transformer.