We have only tested with 3 leaders and 3-6 followers. We already saw that --cluster add-node is difficult to handle in the context of a partially broken cluster.
We should have tests with a different number of leaders to see how multiple leader failures are handled by the operator.