Steam stopped attempting direct P2P and relayed everything
A March Steam Networking change routed every P2P match through Valve's SDR relay, spiking latency for players far from a relay POP until a client fix shipped.
Around 13 March, Street Fighter 6 players in Israel saw PC-to-PC matches climb to about 120ms of ping. The same players against European opponents stayed at 60-80ms, and cross-play from PC to a PS5 held at a clean 5-10ms. Every affected match ran over Steam Networking. Games with their own P2P stack, like Tekken 8, kept working normally.
The reports did not stay local. Players in Egypt described the same pattern, and a separate set of reports from China covered Warhammer: Vermintide 2, where Steam Networking stopped establishing a direct connection at all, even with “Share IP Address” enabled on both the host and the peer. Every session was being relayed through Steam Datagram Relay (SDR). Deep Rock Galactic and other SDR-dependent titles showed the same behavior.
SDR routes game traffic through Valve’s relay network. That buys privacy, since peers never see each other’s IP, and some protection against DoS, at the cost of a hop through the nearest relay point of presence. For a player close to a relay the penalty is small. For players in Israel, Egypt or China, the detour is the difference between a playable and an unplayable match. Older Steam clients used STUN to open a direct path between peers. The newer Networking APIs stopped attempting it and relayed everything.
“Share IP Address” was the control meant to permit a direct connection, and it was being ignored. The P2P documentation notes that on Steam traffic is always relayed and IPs are not shared “between untrusted peers, so that malicious players cannot DoS attack.” One reading floated in the thread is that the change began treating effectively every peer as untrusted. Valve’s engineer confirmed they were working with SDR partners to find out why the toggle was not being honored.
Blocking SDR server traffic to force a direct connection made things worse. Matches failed to start, because the client no longer falls back to P2P when it cannot reach the relay.
The community workaround doubles as a diagnosis. Copying steamwebrtc64.dll from the Steam install directory into the game’s binary folder, on both machines, restored NAT traversal and produced a “NAT traversal successful, IP shared” result. Doing it on one side only cleared the visible 3999 error but still routed the session through SDR. Melty Blood needed the 32-bit steamwebrtc.dll because it links steam_api.dll rather than steam_api64.dll. The January client worked too. Both facts point the same way: the WebRTC and STUN path still existed in the install, the current client had simply stopped calling it.
Pinning that down took about a month. The SF6 players reported first to Capcom, then to Steam support, on the assumption it was a game bug. It was a platform networking change that surfaced in whichever title you happened to play.
If you build on a relay-by-default transport, this is the failure mode to watch. A relay policy change is invisible to your application until latency shows up in the field, and it shows up first for the users farthest from your relay POPs. The opt-in toggle for a direct path stopped working with no error, and the only signal was ping. Monitor real peer-to-peer latency by region instead of trusting that a “share IP” setting still does what it claims.
Valve shipped a fix: first a Steam client beta, which worked as long as both players ran it, then the change landed in the main client. One item stayed open in the thread. On Proton, the fix appeared to do nothing, and the Linux network stack may handle this path differently.
Keep Reading
WindowsMicrosoft's November update broke SLMGR activation, fixed Friday
Microsoft shipped out-of-band updates on Friday to fix Windows activation failures its November 2026 security updates caused for SLMGR users.
supply-chain-attackOn 16 November, Optimizely served malicious JavaScript to 178 sites
On 16 November 2026, attackers injected malware into 178 brands' sites through Optimizely Web, swapping crypto addresses, skimming cards and hitting iOS.
openjdkSixty-four A's and an equals sign break OpenJDK's Base64
Java4Shell (CVE-2026-114938) is a heap overflow in OpenJDK's Base64 decoder that Calif turned into pre-auth RCE across widely deployed Java apps.
Latest on the Wire
Full wire →- AI Coding Agents Boost Output But Code Review Bottlenecks IncreaseHacker News
- AI Overviews Cut Organic Traffic by 40%Hacker News
- Apache Patches Critical Roller RCE VulnerabilityBleepingComputer
- Complete Web Browser Built from Scratch in PythonHacker News
New signal daily · RSS
Stay in the loop
New writing delivered when it's ready. No schedule, no spam.