All of the solutions covered so far are master/slave solutions that, while asynchronous, expect fairly tight contact between all the database nodes. For example, the master in a hot standby deployment will eventually run out of disk space and crash if it can't ship WAL files to the slaves it's feeding.