John, you're absolutely right - it's 1/4 hp, not 1/2. Clearly limited use, and my previous post was intended for bench test, where presenting a load might be challenging, and knowing the cutting forces on threading on a lathe, likely too puny to fit to the lathe. Agreed - silly idea. Even for bench test it could be inappropriate if you end up stalling the motor before the oscillation kicks in. Yeah, bad idea.

3 phase - my only caution is battling the combination of the PID on the LinuxCNC spindle control and the PID on the 3-phase driver, I could never persuade LinuxCNC to quickly setting on an actual speed - from memory maybe 30 seconds to get within 2%, with speed hunting around the desired speed. Annoying, and I expect useless for threading (there is the argument, of course, of a spindle encoder negating this - but it's a less than idea starting point). You could battle/inhibit the PID on LinuxCNC - but I grew too many grey hairs going down that path.

Servo?, that's the route I took, in part to remove the whole 3 phase malarky and to provide the energy to remove the need for the belt speed selector (and bull-gear) by rather over-egging the original 1/3hp motor, but also to provide a much more accurate rotational speed drive - LinuxCNC, for me, providing step/dir control into the spindle. But that's largely supported by the mesa control board as well. I'm not suggesting that it's a solution suited to any/all, but for me it was also in part "because I could".