Very strange — and not full explainable with centrifugal force and direct particle-particle interactions.
I have an idea about it, related to the your about unbalance and "whirling":
when the unbalanced top spins in counter-phase, staying tilted towards the light side, if there are loose balls inside, the balls will go towards the light side, because that side is protruding outwards more than the heavy side, and the centrifugal force will push the balls there.
But, as the balls go in the light side, that side will become the heavy side, so the top will tilt towards the opposite side now.
There should be a feed back relationship between the direction of tilting of the top and the position of the balls, and the balls should stop rolling only when the top becomes balanced.
Something similar could happen in your shape-shifting tops; instead of the rolling balls, there are tilting elements that could tilt less or more depending on their distance from the spin axis and the related amount of centrifugal force; the element that tilts more would extend more outwards, making that side of the top heavier, so the top can self balance, (but only if the top is spinning in counter-phase, otherwise the moving parts could only worsen the unbalance).