Chaos engineering: introducing failure in production without losing your job

getting approval to introduce chaos engineering into our production environment felt like pulling teeth initially. the idea of intentionally breaking things in production sounded crazy to management. we started small, running game days in our staging environment, focusing on non-critical services first. we gradually built confidence by showing the value: uncovering subtle race conditions, validating our monitoring, and improving our runbooks. now, we regularly kill pods, inject latency, and simulate network partitions in production on a scheduled basis. how did others get buy-in for chaos engineering, especially when starting from scratch? what were your first steps and how did you scale your efforts while maintaining trust?