I'll open my yap and guess that the motor is waking up and doesn't know the position of the pole, so it goes to both stops to calibrate itself. I can imagine designing a system like that based on a stepper motor that doesn't have any independent feedback on its current position (e.g. optical encoder), and so must rely on knowing that 0 degrees lies halfway between the two stops.
This would be a bad design because not all
installations have the clearance to go from stop to stop.
I know my motor doesn't work that way because it seems to know where it is when I power it on.
This could be some sort of "limping" mode for when the postion encoder has failed.