To validate if this really is an issue of high volumes of API calls causing the system to run out of sockets, or the upstream rejecting new connections, I’ve created a branch based off 1.44.2, with keepAlive force enabled for all http calls.
Also created a docker image from this branch, available at n8nio/n8n:enable-keepAlive.
If any of you can please switch to this docker image, and let us know if this resolves the issue for you, I can create a proper PR.