Chaos engineering practice: steady-state hypotheses tied to SLIs, blast-radius containment, pre-agreed abort criteria and automated rollback, fault injection with AWS FIS, Chaos Mesh or LitmusChaos, and GameDay facilitation that produces fixes. Use when a failover or redundancy claim has never actually been tested, when planning a GameDay or resilience exercise, or when deciding whether it is safe to inject failure into production and how to bound it.
