Another thought (along the lines of Melodyne) is if you have V-Vocal installed on your system (you've been here a while so maybe you owned a version of Sonar that VV came with) it has some pretty intense controls that can twist up a file... and in different ways than Melodyne can.
It's just a lot less stable so you may want to do it in a completely isolated project the bounce/export and import the results into the main project. Like make a stereo mixdown of your backer tracks so you are dealing with one stereo file, import that and the vocal parts into a new project, use V-Vocal to do your mangling, bounce/render then drag the files back into the main project (to avoid corruption, crashes, conflicts).
Might work fine in the main project anyway but yanno... better not tempt fate.
Cheers.