The questions states that if randomization is permitted, most-powerful alpha sized likelihood ratio tests based on X1,...,Xn have power non-decreasing in n.
I have no idea how to show this. Any help is greatly appreciated.