I wouldn't say there is a consensus necessarily, but in general, there is no way to avoid the direct linear relationship between torque and amperage in direct-drive / built-in motors on machine tool rotary axes. This means if you take a heavy, off-center cut in any CNC machine with a directly driven rotary axis, you will need to run a lot of electric current to the motor to keep it from back driving. Having a disc brake, curvic coupling, or shot pin mechanism allows the machine to take heavy cuts when stationary. Rotating the axis with the brake engaged is only possible on machines that have a ton of torque; it is unusual for the motor on an axis to be able to overpower the brake on that axis, but it would still waste the majority of the electric power going to the motor and generate a lot of heat in the control electronics and in the servo.
Worm gear mechanisms are generally non-backdriveable, but can still move accurately while under heavy torque loads. There is also a mechanism nearly identical to worm gears called roller-cam drives that have the same form factor, but have no backlash and have a lower mechanical gear-down ratio with greater mechanical efficiency. They can be backdriven by cutting forces but are superior to worm gear drives in most ways, except for cost. I don't think it would be possible to have the same type of pivot mechanism that can engage a worm gear work on a roller-cam system, as the rollers and worm must be assembled super accurately and might not be able to "swing" into position. So a roller-cam drive would have to engage the machine spindle by axial movement rather than swinging in radially.
I was recently looking at a quote on a 400mm 5 axis machine; getting glass scales on all 3 linear axes would cost about $7,550 more than base price (~2,500 per scale) and rotary axes would cost between $3,600 and $4,400. They include a 10 arc-second rotary scale on the A axis (trunnion) with the base machine price. To upgrade it to a 5 arc-second scale costs between $230 (Siemens) and $1,150 (Fanuc), with Heidenhain in the middle. They didn't say in the quote who makes the linear and rotary scales, but I've heard that Heidenhain makes all the scales and most machine tool builders just use slightly different cables or whatever to make them work with Fanuc and Siemens controls. The base price of this machine is 89k for the Fanuc version and 92k for the Siemens version, I toured their whole factory and I believe the machine is mechanically (almost) identical to a Doosan DNM 4500, except this builder puts 3 sets of linear roller bearings on the X axis. Very rigid looking machine, it's a shame we can't see these types of prices in the US.
BTW, all of the machines with controlled rotary axes rate their accuracy in arc-seconds. All of the quotes I've gotten tell me the same thing; that worm-gear, roller-cam, and direct-drive rotary axes are all precise to within 20 arc-seconds when brand new and are mainly limited by the encoder fitted to them. Worm gear rotaries have backlash within the 20 arc-second range and this number increases over time as they wear from use. Direct and roller-cam systems have no backlash and don't wear out. If you crash a roller-cam rotary, it will backdrive rather than break or damage the rollers, but you can still damage the main axis bearings if it's a hard crash.
20 arc-seconds corresponds to a rotational inaccuracy of 0.00048" (12 microns) on the surface of a 6 inch (160mm) part. So if you get a machine like the one quoted above with 5 arc-second rotary scales, your best rotary positioning accuracy will be right around 0.0001" on a 6 inch part. With linear scales and an air conditioned shop, that 0.0001" accuracy is not bad for a $100,000 5 axis machine.
).
) for the best repeatable results. This seems pretty common. I recall the Haas UMC discussion were people were talking about having there be scales or not on their rotary axis because they weren't consistent enough using just encoders.
