how do you handle it when a microservice deployment goes wrong, you know, need to roll it back

need some real-world strategies for deployment rollbacks in a microservices setup. we have a bunch of services deploying on their own, and even tho we try for zero-downtime, stuff can go sideways. how does your team handle it? do you script auto-rollbacks using monitoring, wait for someone to jump in, or use special tools that do it for you? id love to hear what worked and what you learned.