While migrating to a new endpoint to verify application availability, a wrong configuration was used to incorrectly determine the unavailability of the application. While the configuration was fixed very quickly, the monitoring system kept using the wrong configuration for several minutes.
The application was actually available for the whole time.