It is the stated fact that this happens in a zero wind hover that is perplexing - the occurence in fwd flight is easily explained by the changes in drag on the advancing and retreating sides as mentioned earlier.
To produce an increase in tq when rolling left (US rotation) the aerodynamic load must increase as drag rises and the Nr tries to decay - therefore the opposite must occur when rolling right.
Any gyroscopic theorists have any data about precession rates with and against the direction of rotation - does a gyro speed up if it precesses against the direction of rotation (spin axis)?
As soon as you remove the advancing and retreating side inequalities from this argument it becomes very confusing - any yaw due to sideslip that might have the effect of increasing or decreasing the Nr would be very small and short-lived in the hover, unless you were prepared to let it sideslip into the ground for the sake of science.