Previous work by us showed that in replica exchange molecular dynamics, exchanges should be attempted extremely often, providing gains in efficiency and no undesired effects. Since that time some questions have been raised about the extendability of these claims to the general case. In this work, we answer this question in two ways. First, we perform a study measuring the effect of exchange attempt frequency in explicit solvent simulations including thousands of atoms. This shows, consistent with the previous assertion, that high exchange attempt frequency allows an optimal rate of exploration of configurational space. Second, we present an explanation of many theoretical and technical pitfalls when implementing replica exchange that cause "improper" exchanges resulting in erroneous data, exacerbated by high exchange attempt frequency.