Some remarks on Wald's sequential probability ratio test