As a workaround, when I really need to, I would do pdf to ebook conversion on calibre (which basically does a reflow), and then upload the converted document to ereader instead of the pdf. This does not work on scanned pdf's, but I guess neither it would work on reflowed documents. For scanned documents, OCR solution (the free ones being tesseract-ocr or its graphical forntends for windows, osx and Linux, or Cuneiform (for Linux)) would be appropriate, but I guess that is too much of a hassle - easier to print the scanned pdf and read on paper
