The related case was discussed in the Dec 5, 2018 Support meeting.

This is a long-standing point of confusion for customers. There are other, related issues in this Epic that likely duplicate this same sentiment. However, it's worth logging this separately so that we make a concerted effort to clarify master/slave topology in a cluster and what the minimum requirements are.

And even though "split-brain" appears to be a term used around distributed systems in the industry, it would be great if we could settle on a less colloquial (or disturbing) term for this scenario.