QUIC P2P Session Survival During NAT Rebinding and Network Path Migration

[Aakash2005GS]

In a P2P system using a signaling server with embedded STUN for NAT hole punching and Quinn (QUIC) for transport: when one peer's network path changes mid-session (e.g., WiFi → Ethernet, so its public IP/port changes), the remote peer's NAT drops packets from the new address because no pinhole exists for it, and the QUIC connection eventually dies from idle timeout. How do we survive path migration without tearing down and rebuilding the session?

[Ralith]

Punch a new hole on the new address? This seems out of scope of QUIC per se. I'm surprised you're getting much mileage out of hole punching at all, given the prevalence of symmetric NATs these days.