Our hub service was restarted, which caused websockets to disconnect. The underlying infrastructure was rebooted as part of an upgrade. Failover was working properly and the services was available immediatly after the disconnect.