Currently, it is possible for parallel heartbeat requests to the server to occur at the same time. We should not be firing any new heartbeat requests before the current request returns, and we definitely should not be trying to establish two separate monitoring connections at the same time. This behavior creates a race conditions for two of the SDAM integration tests ('Network error on Monitor check', 'Network timeout on Monitor check') and is possibly the cause of at least some leaks in NODE-4573, since only one monitoring connection would ultimately be kept track of and closed.

See SDAM spec on the subject: https://github.com/mongodb/specifications/blob/master/source/server-discovery-and-monitoring/server-monitoring.rst#requesting-an-immediate-check