If this current implementation handles 10k req/s comapred to 60k/s on golang side, how can it by design provide availability? I am missing something on how that measurement is done?
Set the timeout on requests to longer times and you will receive the responses, as the nature of how schedulers work in erlang is that the more processes you have to handle (in this case 1 connection = 1 process most probably), the slower all the system becomes.
