To be able to force all the connections going to "Primary" node

This is very important for many apps where causal reads or CCR filters can't help as the apps expect the data reads to be consistent across all the connections not only the "current" connection.

We could use max_slave_connections and allow setting it to ZERO in the readwritesplit router that dictates all the reads and writes to go to "master" while still being able to make use of the automatic failover, and transaction replay capabilities.

This is currently blocking some MaxScale deployments for a project and many others in the future.