A deployment introduced code that degraded a KPI and fixing it may take a week. What should you do?

Reverting the source change and rerunning the deployment pipeline is the most efficient and reliable way to recover from a bad deployment. This approach leverages your existing CI/CD processes, ensuring consistency and automation. Logging in to a server to iterate on a fix locally is inefficient and doesn't guarantee the fix will work in the deployed environment. Manually swapping code on servers is error-prone, not scalable, and bypasses your version control and deployment processes. Changing the instance template and deleting instances to force recreation is a more drastic measure that might cause unnecessary downtime and is less precise than a targeted code revert, especially if the issue is solely code-related and not infrastructure-related.