In Python 3.1, a test fails if it writes to stdout/stderr. It is no more the case in Python 3.2 and 3.3. Why? If we decided to not check for unexpected output in Python 3.2+, the following code should be removed:
Antoine removed that check as part of the -j support, if I recall correctly. The check for unexpected output was part of the support for the transition between the old pre-unittest test suite and the unittest based test suite, and does seem like it is obsolete. Presumably we will visually spot unexpected output during pre-checkin testing.