Using CouchDB on our team has been a pretty good experience; nothing like a nightmare.

We used replication on the native apps for offline mode, kept things simpler with online-only for our web app, and have had a very good time overall.

We are also using JSON schema for our CouchDB documents so we ensure we don't have all kinds of wild / wrong data showing up; if you had really hairy, unstructured data, I can see why a dynamic language might help with that bad situation, but for our use case, we were able to take advantage of the powerful replication built into CouchDB and get solid offline syncing for all of our native apps that also plays nicely with live updates in a web browser.

Did you really find CouchDB that bad to work with? I never in the end settled on either couch or mongo for a project long term but when I setup couch to test it out a few years ago I had it up and running with functional replication quite quickly (i.e., one afternoon) from being a complete novice with the platform. It's built in web panel at the time basically handled everything from basic setup to replication quickly.