Yep. And either the Oracle support is completely incompetent at my workplace region, or the more servers you add to your multimaster replication, the less is your total throughput. Starting at the 1 to 2 change.

I still believe it's a case of the Oracle people being incompetent, and the system technically being able to scale at a rate somewhat larger than 1. But if a potential gain can be acquired for a few million grand, but nobody has the technical capacity to actually acquire it, is the potential actually there?