At 19:40 UTC on June 22nd, we applied a networking configuration change to one of our internal services. This change unexpectedly caused an error within another internal service which caused some customer sites to become unresponsive temporarily. We quickly noticed the impact of the configuration change and reverted the configuration, allowing functionality to be automatically restored. By 20:15 UTC all affected sites were operational once again. We identified several improvements that should prevent such an incident in the future.