Good point. Adding multi-threading while keeping the existing elisp compiler would definitely improve responsiveness, maybe enough so to turn the issues I mentioned into non-issues. That said, the proposed approach has the potential to kill two birds with one stone: add multi-threading AND improve raw performance.
The problem with multi-threading in emacs isn't a lack of libraries to support threads, as CL would supply, but rather the complexities of synchronization of a large existing codebase, which porting to CL would have no effect.
