How should we handle recovery from interrupted operations using the new Automation API? Even in the CLI approach the support to mitigate this automatically is done very poorly.