On the evening of Jul 29 we changed our api configuration. Everything was up and running, but we did not change the monitor, leading to what appeared as 15 hours of downtime despite everything being operational.