Monitoring messages are sent between RS instances every 3 seconds by default. As the number of RS instances grows, so does the amount of background "chatter". We should investigate a) whether we can reduce the default frequency, and/or b) reduce the size of the monitor messages.
Each monitor message may be quite big. In particular, the amount of network traffic could be as bad as O(N^3) with the number of RS instances since each RS is sending monitor messages to all other RS instances (i.e. N(N-1) messages per interval) and each monitor message's size is proportional to the number of DS+RS instances (TBC).