This is helpful with PyMBAR is run on CPUs (which is the only option with PyMBAR 3) and we want to avoid clogging the GPU queue. Discussed in https://github.com/michellab/a3fe/issues/17.