If you save time step files, could you visualize and tell where omega goes far beyond? In this way, you may see where the fault is. I still think, the problem is in BCs.