Change lazy_import to allow proxying when necessary. Write a monkey-patch for epydoc.uid.ObjectUID so that it can understand a lazy_import object. With the monkey-patch in place, epydoc seems to be able to process bzrlib. Now we still need to fix the errors. :)