Thanks — you're asking whether LiveKit open-source supports relaying media across nodes for the same room, or if everyone must connect to one node. Great question. Short answer • LiveKit open-source does support cross-node media relay for the same room. Participants can connect to different nodes (for example, the nearest edge), and media is relayed between nodes so the room is distributed across multiple servers. What that means • A room/session is a logical construct and not tied to a single physical node. • Each node holds local participants and knows about remote participants on other nodes. • Nodes form a distributed mesh and relay media among themselves (server-to-server) so clients don’t all need to connect to one node. How it’s implemented (brief) • Server-to-server relay uses a custom FlatBuffers-based protocol over RTP rather than WebRTC/ICE to keep server interconnects efficient and simple. • Nodes discover each other and build optimal relay paths, which helps with scalability and fault tolerance. If you want next steps, I can help with: • Guidance on deploying a multi-node/multi-region LiveKit topology. • Configuration details needed to enable node discovery and server-to-server links. • Recommendations for testing cross-node media performance in your environment. Which would you like to do next? Sources: How we built a globally distributed mesh network to scale WebRTC | LiveKit | Announcing LiveKit Cloud | LiveKit, with interactive elements

Thanks for calling that out — good question. Short answer • Cross-node media relay is not available in LiveKit open-source v1.9.12. That distributed, multi-home SFU mesh (server-to-server relay between nodes) is a LiveKit Cloud capability. What that means for self-hosting • In OSS v1.9.12, a room/session is hosted by a single SFU node. Participants should connect to the node hosting that room; there’s no built‑in mechanism for nodes to form the Cloud-style inter-node media mesh and relay media across regions. • The core SFU, APIs, and SDKs are the same between OSS and Cloud, so client code is portable — but the Cloud orchestration and cross-node relay are additional infrastructure on top of the OSS SFU. If you’d like, I can: • Help you design a self-hosted topology for lower latency (for example, suggesting strategies for routing users to nearest nodes and colocating rooms), or • Describe the specific Cloud features that enable cross-node relay so you can compare trade-offs. Which would be most helpful? Sources: Announcing LiveKit Cloud | LiveKit, with interactive elements