How to handle sequences and replication?

I'm totally new to distributed databases and rather new to SQL in
general (only worked with it 6-8 months).  I am trying to set up a
fault-tolerant system which will allow clients to operate in a
disconnected mode, using a local database, when the network (and the
primary DB server) is offline, and synchronize changes later.

I know that write replication is part of the answer to this, and that
there are some tools available already to help with postgres write
replication.  However, there are some pretty obvious holes needing
filled (for example, the app depends on integer key columns using values
supplied by a sequence, which seems obviously wrong when there could be
multiple clients operating disconnected servers using the same rules)
and I'm wondering if there are standard approaches to this kind of
problem.

Are there good resources out there to help with the design of
fault-tolerant (i.e. network failure, server failure, client-peer
failure) database apps for networks of clients and servers?  books, web
sites, HOW-TO, etc?

Thanks,
Bill Gribble