I'm also concerned about long term readability. Some of those documents might have at least a few hundred years of people wanting to read them before them. Can any of those innovative new data formats stay around long enough, especially if they're proprietary? We might change computation platform and hardware paradigms a few times as well, just think back to 20 years ago. A lot of measured raw data is already lost due to this; it's about impossible to parse any more. Pure text can survive far more easily.
Well, many of the existing publishers (at least ACS, Nature group, Wiley, Science group) are already pushing hard towards something "more" than PDF, mainly for DRM reasons. And scientists in general are resisting greatly, because PDF is such an established format, supported on all devices and platforms, can be created by the tools we use, and can be shared freely amongst peers. The barrier to adoption for a new format is extremely huge.