If those "talkers" all had hubs which were participants in the mesh, would that help the network "scale"? Also, _what is_ the "scale" issue? Network congestion

If those "talkers" all had hubs which were participants in the mesh, would that help the network "scale"? Also, _what is_ the "scale" issue? Network congestion? Latency introduced by the need to relay a message from node A1 to node Zn? Something else entirely?