The applications using the database were impacted by failover of the load-balancer. We mitigated this short-term by clearing up the stale connections. We will work on the long-term mitigations: review of the connection pooling configuration and improve the connection usage monitoring.