This will act as a safeguard against a split-brain scenario, where a non-leading Master still believes it is the leader.
As an example, this can occur if the disk and network I/O stall on the leading Master. In this event, most components cannot make progress, as they perform logging. However, components that do not perform logging can continue to operate. Of particular interest is the SlaveObserver, which will continue to operate and inform the Master that all slaves are unhealthy due to the network being down on the Master.