Looking at the G-REDL description of roll and yaw, although accidental control surface inputs will also be part of the picture, the motions appear consistent with gyroscopic forces occurring when a downward force is applied at the rear followed by the opposite.
I expect that this is indicative of the supporting structure coming apart piece by piece. We know that the root cause in G-REDL is in the epicyclic but still these forces are consistent with the suspension bar arrangement progressively failing. I conclude that whatever the root cause in a rotor detachment accident there is going to be a confusion of evidence as the forces on the rotorhead rip things apart.