Ahhh... Sorry for running off into the weeds... my mind can be overactive... Told you I was confused
As I mentioned Linear Technology had an opamp circa the mid to late 1990s, that had an app note to bias the output with a FET with about 2 mA to keep the the bottom transistor in the A/B output from turning on and biasing the top transistor into essentially class A... Still trying to find it... Most simple (1980s/1990s vintage) opamps may not be A/B until the last stage... Maybe someone else can chime in with an opinion on that concept as I reserve the right to be wrong...
The only thing I can find is the datasheet for the LT1010 which uses a LM334 for about 2mA bias on the output of the opamp... but the distortion is not as nifty as say the new National modern op amps...
But I wonder if you could run a comparator into linear mode (it can be done with some glue circuitry stuff, but it may be helpful to know how they designed the chip exactly).. My last microchip design experience was with GaAs HBTs, so not FET in that regard, but I was experimenting with SiGe zero threshold FETs ("zero" Vt if you can imagine that, +/- 10mV of course, dancing on that fine line between depletion mode and enhancement mode)....
Notwithstanding, the newer high speed comparators from the likes of Analog Devices are SiGe, but I do not know if they are FET (they might be, but I could be wrong, as it seems most modern chip design at least for the front end is some sort of FET, perhaps with BJT output)... They are designed to be open Drain/Collector or open emitter whatever or for ECL, RTL, CML et cetera logic... low voltage though, but I wonder of the max specs for output stages... There is latching on some of these comparators so maybe this idea is half baked and a no-go?
You can see some of the simplified schematics if they disclose it to get an idea on how to proceed for research purposes.. it's always fun...
Cheers...