Originally, I wasn’t sure how to make the ‘chaos monkey’ idea work. I wanted to bring in the resources and run my test code in the same application, but couldn’t think of a way to make that work. Because after you import a resource, you can’t change it again in the same run. Then I remembered there’s

EditDirs

in

ProgamTest

, which lets you apply multiple steps. So now I’m thinking the way this would work is you take an existing Pulumi application as the first step, then subsequent steps are additive updates to modify the resources. And the tests would occur within each step (or potentially

ProgramTest

itself). But thinking through this, I’m not sure this is actually any kind of improvement over writing directly against the provider APIs. Thoughts on if this ergonomically seems useful? I do like the idea of the fact that with each step, the state of your infrastructure is quite clear (vs writing a bunch of API calls to modify resources).