Yes, that was off. Iām making the publisher node subscribe to each topic. Anyway, with floodPublish it would be the same, just pushing to a different subset of nodes ā¦ but I did not yet implement that part 
I have to organize the code a little bit into meaningful patches, then I will release that as well.
One thing I am checking is whether the numbers are different during a longer experiment, because TCP slow-start might influence my numbers.