Probably the best place to see a list of changes is the new runtests.py file, which includes doctests for everything that wasn't working before and has been fixed, along with one or two remaining unresolved issues. The file can also be executed directly to run the tests.
While working on this library, I've found that the text concatenation approach is really hitting the limits of its usefulness, and that a tree-based approach is the only way to solve the HTML to Textile conversion problem properly. I may start another project with this approach when I have time, and reuse the unit tests, but for now I think I've pushed this approach pretty much as far as it can go.
P.S. It would be great if you could push a new release to PyPi if you decide to accept the pull request.