Hmmm... I can't speak to the new pandas stuff, but the random module in numpy should cover the same ground as most of R's sampling schemes. (http://docs.scipy.org/doc/numpy/reference/routines.random.ht...) It's trivial to use it to pull indices from pandas data frames.
This is obviously mostly a bug fix release, but I'm never unhappy to have bugs fixed. Sample seems like it'll mostly be a handy shortcut for experimenting, since np.random can generally solve those problems. There's also a new BusinessHour feature, which seems pretty nice.