Yeah, for programs that run in a once-and-done kinda way your approach totally makes sense. This is intended to be embedded into production environments where m

Yeah, for programs that run in a once-and-done kinda way your approach totally makes sense. This is intended to be embedded into production environments where many different types of logs could spew many different kinds of multi-line information and there needs to be an efficient way to capture all that and send it for automatic diagnosis and potential further manual/automated actions (if applicable). I thought about it from a cloud deployment (Kubernetes) perspective (like a node DaemonSet).