The fact that disable_master_failback does not work with root_node_as_master is not documented

The fact that disable_master_failback goes agains what root_node_as_master tries to accomplish is not documented and the two parameters can be used together without any warnings.

Original title: Maxscale | Multiple instances in a cluster go out of sync.

Original description:
Hi there,

We have multiple instances of Maxscale running in the same DB cluster. Occasionally the maxctrl administrative client reports an inconsistency in states amongst the nodes, e.g. Maxscale #1 reports server #1 as the Master, while Maxscale #2 & #3 reports server #2 as Master etc... . Could you suggest the steps to correct these inconsistencies?

Thanks.