The strategy followed by pypy to reimplement numpy from scratch makes it rather unlikely that it will support packages depending on numpy, because of the various dependencies on numpy's implementation details.
That reinforces the goal of porting NumPy (and PyPy's trackrecord at achieving such ports). That is all the people not using pypy cause it lacks numpy will, after this port, have the option to.
It's not possible to do cool stuff with reusing - like parallelizing expressions etc. The architecture as it is now already can score 2x wins over original numpy with array expressions and we expect it to get only better with SSE and more parallelizing. This requires reimplementing numpy.
This is again a different argument. I understand it is more fun, more rewarding and more challenging to implement a new array module on top of pypy. But it is seriously doubtful that it is the best way forward to make pypy usable for libraries which depend on numpy.