Having looked again at the diagrams, it appears to me that the castellated nut assembly should be locked on the threaded end at the end of the shaft and it should be able to freely rotate with it. The duplex bearing should allow this. However, the duplex bearing seized, or partially seized and produced an extreme overheat. This resulted in the locking nut welding to the carrier and it then it unscrewed as the shaft continued to rotate. The carrier, which is vital for the pilot to input yaw demands, was then no longer connected.