Which is why ePub and other digital book formats exist. I'm glad there is a format that prioritizes WYSIWYG over reflow (though the part of the spec where the introduced scripting is a bit dodgy)
Should be able to easily reflow text as long as it's using newline operators (T*, ', "). Might still need some basic heuristics for paragraph breaks. But much better than the alternative of attempting to correlate individual lines of text together based on positioning.
I use calibri for my E-Books, and it seems to do a pretty good job of reflowing PDFs(and exporting to mobi). YMMV though, especially if headers/footers are badly done.