loadbalance distribution ensuring repartition after failover

Since 3.1.0, load balance is round robin. Problem is that when a server fails, the repartition is then distributed on other servers. when the failing server is up again, the load will take some time to come back to normal : the distribution will not be distributed immediately,

Goal is that after a failover, that connector priorize connection to the server that doesn't have as many connection, in order to have a good repartition.