I am sure it is not, unless the motor happens to perform significantly worse under light load.
What happens is this: a belt drive is primarily a reactive (resonant) system formed by the inertia of the platter bouncing on the compliance of the belt. This system is lightly damped (only by the belt's losses) and thus of fairly high Q. It is also quite non-linear. So speed is expected to be modulated at the drive system resonance, its lower harmonics, the suspension system resonance (if any), and any motor breakthrough. Most of these are high Q. Adding damping in the shape of drag (i.e. a speed dependent force, such as viscosity, and not just a static force!) knocks down these modulation peaks.
I ran some simulations of this years ago, but it taught me that one needs quite a lot of viscosity if it is applied at the main bearing. An eddy current brake at the platter's perimeter seems to be more efficient. Oh, and when active controlled can be used for dynamic speed control, within limits of system bandwidth.