What is happening is a function of angle of attack, sideslip angle, the angular body rates of pitch, roll and yaw and the moments of inertia. Without a whole lot of equations things like roll rate coupled through an angle of attack can couple into pitch, etc. With the right combination of static aero forces, dynamic derivatives, moments of inertia and angle of attack and sideslip you can get the hesitation in what would otherwise be a steady rotation.

The simplest I have seen is a high angle of attack roll in the F-15. Under some conditions the roll rate starts out nicely big, the sideslip angle build us and the resulting roll due to sideslip actually slows the roll rate and can stop it. Of course a little rudder in the right direction stops the sideslip and the roll is nice.

In spin tests you can see the oscillatory nature of the rates and angles. At times with some airplanes the can become large enough to actually "toss" the airplane out of the spin mode. There probably is a series of aerodynamic force to inertia force relationships that would enable you to predict when this would happen and would be very important when spending millions of dollars.

On our models it would be interesting to tape on some vertical tail area or vary the moments of inertia with wing tip weights to see the airplane response to the variations of moments of inertia, etc. Of course its much easier with a six-degree-of-freedom program and wind tunnel to 90degrees angle of attack and sideslip angle.