In the future, maintenance like this shouldn’t require downtime. It was necessary in this case because maintenance needed to be performed directly on the production instance as normal deployments don’t work properly. See https://github.com/radian-software/riju/issues/168.