> This lets us avoid single-host bottlenecks, effectively using the whole mesh as a distributed cluster for message forwarding. (Cite scalability numbers here.)
In case someone that can fix this is reading here
In case someone that can fix this is reading here