Langevin Monte Carlo for strongly log-concave distributions: Randomized midpoint revisited