Agree with ta0 (1) that the frame's air resistance contibutes significantly to the total braking torque acting on the skinless top at speed, and (2) that a smooth skin should be a big improvement on that front. Hopefully enough of a torque reduction to offset the increase in bearing torque from the added skin weight.
However, drag by itself would never cause the halting bumblebee descent observed. And neither would simple sliding friction at the tip.
So something in the system must be binding intermittently. If not in the cable run or the bumblebee wheels, I'm with you in pointing the finger at your tip pivot bearing.
Assuming you want the resting top to at least hint of its willingness to fall without the tether at the crown. If so, your pivot bearing has to support a shifting 800+ kg load while allowing at least a few degrees of tilt below critical speed. Yikes!
Have you considered a ball tip resting in a shallow socket supported by a roller bearing? Might reduce tip resistance without keeping the top from falling over. Works well in LEGO tops and might scale up to your case.