It would be nice to experiment with variant package/symbol name resolution. We have package-local nicknames as an extension, hooked in using an undocumented find-package-using-package function; we probably need to make that generic and complete the work on subclassable structures, then develop a protocol and hook it all up.