This conversation as well as this one suggests that it is sometimes not possible to establish a connection to the "ordinary" UNIX socket at e.g. /tmp/.X11-unix/X0 or wherever the X server listens.

In the above cases, it turned out that the socket name wasn't wrong, but that the socket was of the abstract type, not the pathname type.

Maybe clx should try opening the socket using the pathname type and fall back to trying the abstract type. For SBCL, this translates to (make-instance 'sb-bsd-sockets:local-socket) and (make-instance 'sb-bsd-sockets:local-abstract-socket) respectively. I'm sure other implementations support this as well (if nothing else, it should be possible to stick a #\0 into the path as described in the man-page).