You bring up a good point about the mechanical system - a worm drive is superior for the application for the reasons you stated.
I have to disagree with what you're saying about the stepper motors though. 400 steps/rev accuracy? No, the stepper motors are essentially servo motors with a 1000 line encoder - after quadrature that's 4000 counts/motor revolution - that's 0.09° resolution. Accuracy is a function of the drive system and tolerance on the encoder. They will have 12Nm holding torque, not 6Nm. Holding torque is always higher in a stepper motor than dynamic torque and typically what the motor is rated for.
Hanermo, the motors linked provided the torque curves over the operating range. What you said may apply to the junk high inductance chinese stepper motors, but these will reach 1000RPM (at least the 8.0Nm one will).
Schneider/IMS made a Nema42 sized stepper that will run up to 3000RPM
http://motion.schneider-electric.com...s/MDM42_AC.pdf
Some small stepper motors can hit 6000RPM.
Your real torque depends on the torque curve, not a thumb calculation of the nameplate - that's only something you assume when you are dealing with steppers that have no listed specifications.