Never having been on the 'other side', I obviously have no idea about the inner workings of the publishing process. I know however that journals published by the American Institute of Physics (AIP) are incredibly streamlined when it comes to LaTeX. They have superb templates which allows one to format the article such that it looks pretty much like the final version. No messy conversions and danger of mistyped equations anywhere. Wouldn't it be possible for PLoS to tap into whatever knowledge/resources AIP has? Maybe the answer is no, due to incompatible systems -- but it might still be worth exploring. (I really want to send more articles to PLoS but I also really don't want to go through the messy conversion process again...)