Workers and queues fail. SQS was down for us for almost two weeks while AWS fixed a bug. We had no choice but to wait or rewrite our implementation... Again! We've already had to rewrite once due to poor visibility and rare occasional problems processing data. Debugging such distributed systems is legendary hell. And that's just for simple async processing so that we can return a response quickly to the user and finish the task in a few seconds. There is simply no comparison between such a complex, failure prone distributed system and the simplicity, reliability, and ease of use of having support built into the language for this, IMO.