What cloud change would you want to test before it ships?

A couple of recent conversations have shaped what we’re building in Cloud World Model. One led us to model how provider API limits affect bulk updates. Another focused on whether workloads could recover within an EKS Spot interruption window.

Both came down to the same question: “What happens if we make this change?”

I’d like to hear a real example from your work. Was there a cloud change where testing cost, capacity, rollout timing, or failure behavior beforehand would have helped you make the call?

And what evidence would you need before trusting a simulated result enough to act on it?