There are users that have specific workloads where iproto thread is the bottleneck of throughput: iproto thread's code is 100% loaded while TX thread's core is not.
For such cases it would be nice to have a capability to create several iproto threads. I think we should share listen socket(s) between network thread while every transport socket should be managed by one of network threads exclusively.
Of course there must be an option like iproto_threads that haven't to be dynamic.