Database timeout connection is not properly handled when connected to Galera Cluster

BTW note that a few days ago I increased the timeout to 7400s on the load balancers. Before doing this, the fts3rest_error log was constantly full of errors about connections to the database being dropped. The database timeout for the rest api seems to be 7200s, and the default load balancer timeout was < 1 minute, so I assume that the load balancer was timing out connections but the sqlalchemy stuff wasn't aware of this, then things failed when they tried to use the connection to the database.