My team has also become good at that. We do "game days" where we push a simulated breakage and the team tries to fix it, but after a few "faulty" code pushes, everyone checks the rollout history for recent code or configuration changes and suggests a rollback as the first step. Having this log makes it very hard to break things!
This just means the breakers have to be more clever. My favorites are taking a global lock in Postgres (all reads and writes instantly stop, and someone has to figure out how to kill the session that holds the lock), and installing a service worker that blocked traffic to our domain. (You can roll back, but the service worker is still there!)