I would love to understand your experiences. How do you detect issues? Are you using feature flags? Have you had any disasters? How do you deal with problems? What are the biggest pain points? Kubernetes? Etc?
What's there to know? You divert some orthogonalizable flow (orders,
accounts, widgets, etc.) to a parallel setup. The trick is orthogonalizable
because if something in the main flow depends on something in the test flow,
you're in for it.
But how do you detect when something is broken or know when it is good enough to deploy everywhere?
How do you know something isn't broken in production?
At first I interpreted this as a spin on "everyone has a test server; some of us are lucky enough for it not to also be prod."