Yes, internally Pandas stores the data as a series of homogeneous arrays, which correspond to one more columns in the data-frame. Details here: http://www.jeffreytratner.com/slides/pandas-under-the-hood-p...

I agree with what you say except that I consider data-frames one of R's strengths. What makes R data-frames great is that the language is designed around these data structures, thus allowing most of their inherent limitations to be overcome by following "good practices". The problem of porting data-frames to other environments as in the case of pandas in my opinion is precisely a lack of language support, which makes the whole thing feel a little stitched together.