We deployed a new version of a proxy system in the VTEX IO platform that caused the admin environment to go down. We rolled this deploy back in the first 10 minutes, but it did not fix the problem. We tried switching the traffic to our staging environment, which is used by all VTEX employees as a way of testing new versions, while we fixed the production one. This did not produce the outcome we expected because the staging environment is less resilient and also went down due to the sudden load increase. The change, however, alleviated the production environment load, allowing it to recover and the admin to start working again when we switched back.