"with modern machining bearing tolerances can be very close to perfect"
Agreed. It’s the "close to" that tells the story. In the old days, bearings were machined fairly closely, and then rotated against each other in situ, with fine grit, to smooth them. The finest bearings were done with the finest grit. That’s exactly how the glass-stoppered reagent flasks from the chemistry lab were made - and they were literally air tight.
That’s also similar to how the billion dollar telescopes are made today, so I doubt that the practice can be improved upon with machining. With an air cushion, however ...
Thanks for your explanation of Raul’s point - it was opaque to me. To that point, a platter bouncing on an air column is clearly visible and audible. It can be tuned away by changing the pressure by a few psi. Ask me how I know.
Heavier than 20 lb is absolutely useful for inertia when the stylus is dragged through the groove, producing a variable retarding torque. It is also useful for resisting the irregular pulses of an electric motor, and the bearing noise in that motor.
I have tested this - with a 45kg platter, the precision 2W electric motor, run from a quadrature power supply, generates enough noise from these sources to be repeatably audible, as a slight HF glare of the CD kind. Removing the belt tells the story.