We should investigate whether or not error messages are sensible when forwarding Unix sockets e.g. from Windows workspaces, and whether or not the behavior is the same between Windows -> Windows and Windows -> Linux.

And, not for this PR, but I know that we do have customers that target Windows workspaces, and we should verify we get some sensible error if you try to remote-forward a Unix socket from one.

Originally posted by @spikecurtis in #11631 (comment)