Everyone who says that ANU input on the tumb switches will reverse MCAS is right.
However, the question is whether training was remotely adequate to address the AND authority of MCAS, which demanded an unusual amount of ANU trim in the teeth of stick-shake, AoA warnings and UAS advisory. Anyone is free to correct me, but it seems as if an unusual amount of thumb-switch ANU would be needed.
Meanwhile, the intermittency of MCAS inputs seems to be tailor-made to look unlike a runaway. It certainly doesn't serve to decrease confusion.