Thanks for the quick thorough reply, and from the creator of Phoenix no less!
Sounds like I made a few assumptions about Phoenix.Pubsub and you know what they say about that.
Correct me if I’m wrong: subs are stored locally only, and messages that are broadcasted via pubsub are always sent to all nodes, correct?
If this is the case, that seems like you’d move the bottleneck (or at least more load) to the broadcast method, no?
The bench is cool and quite promising. However, it takes place all on one node, so it doesn’t test the distribution of subscriptions. It would be super cool to see a bench that tests distributed subs. (Perhaps I can help) The link I supplied shows a very steep drop off of subs/sec the more nodes you add for most of the tests he did.
Sorry if what I’m saying sounds overly negative, I’m super impressed with what’s been accomplished here and think that this is great for most use cases. I just like the concept of totally maxing out the scalability and have a feeling many people in this community feel the same way 
I wonder if there could be a pluggable eventual-consistency based pubsub backend. I wouldn’t be opposed to working on that myself, if people think that could be a good idea.
