It would be far better to replace refereeing with TruthSift.
Depending on the desires of the Journal, the topic could be public, or private restricted to the author and the referees and editors.
The paper would be downloadable by referees using a link at the topic page, which would otherwise state "this paper should be published as is.". The referees (and if it was desired to allow it to be public, anyone else) would then add proofs and refutations to the topic statement. The author could rebut statements, possibly after making edits.