The least that comes to mind, is releasing a second counterweight in the same end as the load. The counterweight is allowed to shift to a higher radius, giving the same moment as before. This keeps it statically balanced. The dynamic imbalance lasts between the instant the load is released, and the counterweight smacking into its final position. Which should be about a quarter rotation, I would guess. The catch (literally) is, the bearings still need to withstand that 1/4 cycle of massive imbalance, and, the counterweight will be accelerated (radially, i.e. in the rotating frame) very rapidly, then decelerated EXTREMELY rapidly as it catches the end stop. This can all be designed for, but a good question is which is harder: doing a supersonic baseball pitch and catch in the span of a few feet, or just letting the fuckin' thing bang itself around for a few -- whatever it takes to servo it more slowly, or brake the rotor down to safe levels -- maybe tens, hundreds, thousands of cycles?
I suppose these are even high enough forces that we'd have to consider the flex of the shaft, even as thick as it is, with respect to the clearances of the bearing, and whatever bearing surfaces there are. What would a high speed vacuum bearing look like anyway, would it be a simple hydrodynamic (oiled) journal bearing? Ball bearings surely won't be used here; roller bearings perhaps. Some kind of lubrication is still needed, I think? (Vacuum bearings aren't new for aerospace; they're well understood...
at least, sometimes.) I don't know offhand what they use for these things; some kind of vacuum grease would be necessary to avoid loss and contamination (spreading / condensation of oils on all surfaces in the sealed environment), but at the same time, vacuum greases are generally quite goopy. And graphite and MoS2 work by basically grinding themselves up, and in a completely dry environment would make more dust and thus contamination. So, I don't know, but it's a thing that's been done, however they've done it.
Tim