I asked in Python-Dev, and if I understood the answers correctly, we can make changes to 2.7 and 3.2 without fear. For 3.3 risk is higher, so we need to be more attentive. However patch for 3.3 contains not only the potential improvements for other tests, but also bug fixes (I think these bugs should cause tests failure on the platforms where sizeof(Py_ssize_t) != sizeof(long) or sizeof(Py_ssize_t) != sizeof(void*)).

Of cause, if Georg Brandl not vetoes this.