Investigate and fix peers frequently ending up on different rounds of consensus

[evan-forbes]

Block producers are producing block consistently on the testnets, but we aren't doing so at a reliable rate, or within the first round of consensus #939. While its not a blocking issue, in that the network isn't halting, we are consistently seeing error logs of peers that are on different rounds or heights. We should investigate this issue and implement any fixes (assuming that we end up needing one).

[evan-forbes]

ref #965

[evan-forbes]

on mocha using the latest params and block sizes, we see the same percent of heights that require more than one round to reach consensus as networks such as gaia and osmosis. Therefore, I think this issue is now safe to close.