Right now the extra_port defined for a server will only be used when a connect attempt on the regular port fails

"This allows MaxScale to connect even when max_connections has been reached on the backend server. If this parameter is defined and a connection to the normal port fails, the alternative port is used."

As even when successfully connected to the regular port monitoring queries on that connection can stall while waiting in the pool-of-threads execution queue it would IMHO make more sense to have the monitor(s) use the extra port by default, to ensure that monitoring queries and switchover/failover actions are handled with priority and don't have to wait in the thread pool execution queue if the thread pool is fully loaded with query execution threads already.

Monitoring queries will still suffer from CPU overload if pool-of-threads was not properly configured to prevent this, but at least they won't additionally suffer from having to wait in the thread pool queue in such cases, too.