The formula used for setting the initial histogram size in AWH uses a Gaussian geometry factor, taken from Metadynamics, and sums inverse crossing times. The Gaussian geometry factor is actually not relevant for AWH, so should be removed. For multiple dimensions we should sum the crossing times, not the inverse times. For 1D this changes the initial histogram size by about a factor of 3, which doesn't affect the convergence much. For multiple dimensions with different covering times it might be affected more. We don't want to change this behavior in a bug fix release, but we should fix this for the 2021 release.