Redoing the classic PIMETA design, eh? Both parts are quite decent choices, but obviously the devil is in the details and implementation.

The main drawback of this specific buffer is the quite large voltage drop if large amounts of current are sourced - so do remember that this won't really work to source 250 mA if you give it 3.7V or 5V supply. This is typically not an issue in headphone loads, as high current low impedance headphones are usually sensitive enough. Do note that the result will be current mode drive - balanced armatures respond differently to voltage and current drive and voltage is the more common mode in headphone amps.

Note that V/I converters need a matching I/V converter and amplifier. It makes no sense to use these for short cable runs - they're made for PA systems.

Finally soldered the suckers in! Ended up using the same heatsinks the MOSFETs used.

All in all I like it better than the no-buffer solution. There most certainly is more bass and stereopanoramics have improved as well.

I still can't get my head around why would Parasound ruin a clean opamp output with such a crude buffering solution.

you most certainly did.

Btw, it was your pals' Amos's old DAC. Now it's dead-ish, but your latest findings make me want try bringing it back into the realm of living. ...all the hours I've spent peering into specsheets and schematics, makes me shudder.