> OT has a negative feedback wire built in. It makes sense, but I've never seen it.
Makes a lot of sense.
That's a LOT of output taps. Do you know what they all do? Are any useful?
When you see super-duper output like that, suspect it may be marketed to "special uses". Like a school with a mile of wiring in the walls. You really want that wiring floating and balanced, but the NFB wants to be chassis/ground referenced. A few turns of skimpy wire is the way to go. Fairly common on PA amps. Using the speaker winding for NFB is a bit cheaper, and also necessary for VERY impressive Damping Factor, but that's a hi-fi fetish.
> It also has a non-adjustable fixed bias supply. Isn't that how Mesa does theirs? Seems a bit silly
If you know your tubes are consistent it works very well. Especially if you bias to the cool side (don't worry over-much about soft-signal rasp).
> I'm thinking of building something simple with two EF86s
I don't see what's wrong with going into the Mixer grid? You have your EF86 Pentode gain-stage, a Volume control, then the ever-popular James tone-stack (here split into two stages for slightly less interaction). Then the mild-gain 6FQ7/6SN7 driver with NFB, and a cathodyne phase-spitter which always works well. ---EDIT...no volume control.....
One design weakness. The common cathode-cap on the two tone-control triode amps MUST be large and healthy. Maybe NBD with the mild 6SN7 plus high-knob loss. I'd be inclined to split pins 3 and 8, use separate 3K+22uFd on each.
Screen resistor is quite large. This may help limit the damage when idiot users abuse the power. If you scale-back the 6k2 to typical g-amp values, you will need more G1 bias but as you say it is un-trimmable. Personally I've had good sound from "large" screen resistor so I'd leave it the way Webster has it.