Why we rehearse rollbacks before teaching blue-green releases

Release training often jumps to shiny deployment topologies while skipping the awkward conversations about who owns rollback decisions. We flip the order: cohorts draft a rollback sheet, assign owners, and only then wire blue-green automation that respects those roles.

Rehearsals stay lightweight. Teams run tabletop exercises with sticky notes for dependencies, then translate the narrative into a checklist stored beside the pipeline YAML. The goal is shared language, not perfect automation on day one.

Students capture telemetry gaps discovered during drills. If nobody notices latency regressions until a customer does, we treat that as a curriculum signal rather than a personal slip.

By the end, each participant leaves with a one-page runbook they could hand to an on-call partner without embarrassment. That artifact matters more than any single Terraform snippet.