As we all know, embedded databases out-perform server databases on every occasion. The problem with embedded databases is that theyre difficult to scale. If I were to create an entire enterprise web application within a single jar file using embedded tomcat and embedded objectdb I would run into problems trying to cluster the data.
It would be really useful if some kind of new external replicator application was created that could be used by server administrators to ensure replication is optimal and healthy.
I believe that there are some open source solutions like Derby that have something in place for this so it shouldn't be too difficult to reproduce.