It's a different tradeoff. Systems that fails and *executes recovery path* from time to time tend to be more resilient to catastrophic failures. Ideally you wo

It's a different tradeoff. Systems that fails and *executes recovery path* from time to time tend to be more resilient to catastrophic failures. Ideally you would want the combination of both: system doesn't usually crash/fail, but you force it to do so regularly. See https://en.wikipedia.org/wiki/Chaos_engineering https://en.wikipedia.org/wiki/Chaos_engineering