We pushed a bad configuration to the public reverse proxy causing some downtime. The changes were reverted once we noticed the outage.