This problem is still occurring. Sometimes it happens, sometimes not. Killing the underlying Lisp process (sbcl) unfreezes Emacs. Here is the error message that appears in Emacs after killing SBCL in the buffer *slime-error*:
it's not necessarily what happens to you, but just in case...
SBCL has a *world-lock*, which is grabbed by with-compilation-unit, which is used by an around method on asdf:operate on asdf:operation (so, on any operation).
in the real world for example: doing a load-system, meeting the debugger due to some compile error, going to the slime REPL, initiating completion (so, in a different thread than the initial load-system), hitting a method cache miss, and finally compiling a method cache lambda... will end up in a deadlock.
you can break out emacs with C-g, and when you cancel or finish the load-system on the lisp side, then your stuck slime worker threads will also go away.