Belts are made circular. When they are installed on a pair of pulleys, they are stretched out of the circular shape and into the "racetrack" shape.
The simulation for this model assumes a belt was manufactured in this shape:
If you turn the pulley in this simulation, a section of material that was manufactured curved has to be bent straight and a section that was manufactured straight has to be bent. You will see the curved sections of belt be transported between each pulley. That will be slightly amusing but will not reflect real world behavior. The state of stress of the belt will depend on the angle of the pulleys.
The proper way to simulate the behavior of a circular belt is to take 2 pulleys that are tangent to the circular belt and move one pulley until it stretches the belt into a racetrack shape. Only then do you want to rotate one of the pulleys to move the belt. In that way, the state of stress of the belt will be independent of the angle of the pulleys.
An exception to this logic is if the web is thin so it can be assumed to have no bending stiffness and only in-plane (membrane) stiffness. Then you can draw the web in the shape of a racetrack, turn a pulley and the state of stress in the web will be the same at any pulley angle.
How do you want to proceed?