Due to the apparent age of the testing and changes to Python string library code since then, this should be ignored until you rerun the tests using the Python version you are using for your application.

In particular, the old assumption that a += b is _horribly_ inefficient is no longer true.