SERVER-126965 was closed out and gave the LogServerTopology atomicity by wrapping the topology in synchronized_value. This gives the topology atomicity, but does not give it thread safety. This is because a flow of code may have the topology change between multiple accesses.

One plausible path to give the topology thread safety is to take copies of the topology and pass these copies to the different callers to ensure that there is a consistent view. In the case that the topology is now stale, then we should retry (there should already be retry logic in place in the log server manager) and then take another copy of the topology with the changes.