In some sense, it is taking "fail fast" and moving the failure closer in time by just imagining the failure rather than waiting for the real failure. This allow

In some sense, it is taking "fail fast" and moving the failure closer in time by just imagining the failure rather than waiting for the real failure. This allows for a quicker feedback loop. It also allows for having the feedback without actually having failure. I suspect the second feature is more important than the first. Feedback from real failure remains more valuable than feedback from imagined scenarios. Thing is, real failure can also be quite expensive.