Anyone seen LiveKit agent workers crash from rapid connect/disconnect spam? Hey, quick question for anyone who has built voice assistants with LiveKit. My teammate was stress-testing a voice assistant we built at our startup. They repeatedly connected and disconnected in rapid succession (every few seconds), and this ended up crashing the agent worker on the backend. Has anyone run into something similar? I’m trying to figure out whether: • this is expected behavior under rapid session churn • I’m missing some cleanup/cancellation in my worker • or if there’s a recommended pattern for debouncing / isolating sessions If you’ve dealt with this before, would love to hear how you handled: • leftover async tasks • race conditions during teardown • preventing old sessions from hitting a new worker • or stabilizing the worker under heavy churn Any pointers, examples, or “here’s what we did” would be super helpful 🙏