For the CGKA sync we can use our old friend RIBLT sync to sync the set of CGKA ops, but for the document content we need to do something a bit different because we want to be able to take advantage of the bandwidth gains we get from compacting Automerge documents. Sync out of sync documents, for each document which is out of date Run RIBLT sync on the CGKA ops Download CGKA ops we are missing UPload CGKA ops the remote is missing Run sedimentree sync on the document content In the common case the server will be able to decode these symbols (because only one document has changed) and immediately determine which document has changed, then the server can send back a response with the sedimentree summary for the changed document and the first 5 symbols of the server CGKA RIBLT state.