The concept with many servers placed in geographically different locations brings some serious problems. To keep the simulation running, all participating servers must be up. If any server broke down, the others shouldn't be able to continue, because the world would immediately become inconsistent and this inconsistency could not be fixed another way than by restarting all the servers from the last consistent backup.

Note that no assumptions about the data service have been made; its semantics and usage is defined by the application using the Core and thus we cannot tell (from the Core's point of view) to what extent a failure of the data service matters or not.