Due to a caching server being too small, we had to manually upgrade a part of the infrastructure. Because of the severity, we could not prepare this beforehand, resulting in a 4 minute downtime for logged-in users. text production and data updates were not affected.