Create a new pull CBLReplication and set its documentIDs property to the document(s) you want to re-fetch.
(Actually there might be a problem with this the second time you try it — the first time you run this replication it will fetch the docs, but then it will remember a checkpoint at the current sequence of the remote db. The second time you run it, it will only look at revisions newer than that checkpoint, so if those docs haven’t changed they won’t be replicated. Hm. What we need is an API to reset a replication by clearing its checkpoint.
A workaround is to add a bogus random doc ID to the array of doc IDs to fetch. This will cause the replication to have a different checkpoint ID every time.)
—Jens