Dear Augustin,
I had a look at this table: http://wiki.postgresql.org/wiki/Replication%2C_Clustering%2C_and_Connection_Pooling
It seems pgpool could be (COULD BE) the way to go.
On top of that by using a middleware in between the client and the server(s) you don't need to change your client code.
I was just wondering if somebody had already used pgpool and with which results...
Thanks,
Maurizio