I think the easiest solution is to have a single ACS that is populated via RDBMS Sync. This ACS becomes the replication "master" that then pushes its config down to a set of "slaves".
That is the easiest method but replication is a destructive write onto the slave - so you may choose not to do this.
An alternative is to use the Sync Partners config (part of RDBMS Sync) which attemtps to process actions in the sync table on multiple ACSs. For this to work you need the "other" ACSs to have the RDBMS Sync'ing ACS server in their network config db.
You need to make sure that ACS can write to the transaction table too (note CSV datasources no good) in case one of the other ACSs is down.
If you're having problems check the rdbms sync CSV & service log on the "master" ACS and the csauth service log on the "slave" for errors.