Hi,
First of all, this sounds a great project. But I think there is a lot of feature that can make it key inside an infrastructure.
For instance, the failover, as a central point of connection, it should be possible to give a list of servers that can serve as replacement if the main one is not responding. For instance a master postgres and a slave are in sync and the master crashes. This failover mechanism will allow to have an application working in degraded mode hitting the slave in read only. We can also imagine that all write requests are kept and replayed once the master is back (this is a tricky point and needs a lot of discussion, limits...)