It's great to have everything-as-code because of reproducibility, etc., except now the cloud infra you are targeting can and will randomly fail in unpredictable ways.

@buherator yup, I’m experiencing it and it hurts badly. AWS CDK can fail in many way one might not find out in a life. A slight mistake in refactor might push stacks into limbo. Also, there are unknown regions of CloudFormation. Even an easy resource change could halt a pipeline deployment because it defies any further mods. Manual deletion of stacks and further retries might help. I said might.