Unstable connections on some Windows 11 systems with KB5063878 or KB5063875 installed

[ramon-juicelabs]

Has anyone seen unstable connections on Windows 11 with either KB5063878 or KB5063875 installed? We are having problems running workloads connecting locally via localhost or remotely in different ways.

At first the issue appeared during the initial connection/handshake phase. Our software is using MSQUIC 2.5.0, with OpenSSL, and its integration is very standard but makes use of client certificates. Debugging the issue showed that it happened during the client certificate exchange and the server never receiving the data and eventually timing out the connection.

I have attached a ETW Full.Verbose trace for a run of the above @ msquic-handshake-timeout.zip There are a number of calls to WSASendMsg which are failing with error code 10057, for example:

  577 [9]62E0.4744::2025/08/24-22:52:55.390930800 [Microsoft-Quic][data][0x1CDBD1BBAB0] Send 1969 bytes in 1 buffers (segment=1220) Dst=[::ffff:127.0.0.1]:63474 Src=127.0.0.1:63475
  578 [9]62E0.4744::2025/08/24-22:52:55.390940000 [Microsoft-Windows-Winsock-AFD]sendmsg: 0: Process 0xFFFFE389022E5080, Endpoint 0xFFFFE38902516AB0, Buffer Count 0, Buffer 0x0, Length 0, Seq 3043, Status 0xC0000140(NT=The connection handle given to the transport was invalid.)
  579 [9]62E0.4744::2025/08/24-22:52:55.390941800 [Microsoft-Quic][data][0x1CDBD1BBAB0] ERROR, 10057, WSASendMsg completion.

Uninstalling the above KB's and re-running the workloads, everything appears to work fine.

To remove any errors with our software and integration we adjusted the sample.c provided in the repository using the latest MSQUIC version (with OpenSSL) to just ping/pong a larger amount of data between the server and client. On the particular host above with KB5063878 installed and connecting via localhost the server after the handshake has problems receiving any data and eventually times out - ETW trace @
sample-trace.zip and again uninstalling KB5063878 fixes the issue.

We have also had reports from users where the connection works but is very slow, this includes localhost connections as well. I'll try and get ETW trace logs for that host as well

Initially we thought it might be related to this problem https://windowsforum.com/threads/ndi-streams-break-after-august-windows-update-switch-receive-mode-now.378629/ but unsure if it is related.