Inconsistent revision hash length

Looking at recent raw documents, it looks like the revision id length varies depending on whether it was generated on Sync Gateway or Couchbase Lite.

On the one hand, these are expected to be opaque, so it shouldn't matter; however the general expectation is that the same change to a revision will generate the same child hash in every location (i.e. if I make the same change in two places, the revision will be the same). Seems like not doing so will generate unnecessary conflicts?

Could be that this is intentional, but if not it's something that we should potentially look into.