For public data, you can try to partition it in such a way that writes can be merged without any potential conflicts. E.g. a user's posts are in a separate part

For public data, you can try to partition it in such a way that writes can be merged without any potential conflicts. E.g. a user's posts are in a separate partition. I have never done this with CouchDB, but the technique is described in Martin Kleppman's __Designing Data Intensive Applications__.