Maybe it's just conservation of angular momentum: if you look closely, you can see that the spin axis and the way she applies the force aren't perfectly tangential to the object's axis, so the lower point of the T-shape is "bouncing" between many positions. If I'm not mistaken, this is explained by the "intermediate axis theorem".