Diagnose API TCPIP Connection Resets
I have observed Ava API behavior that I believe that is abnormal. As the number of simultaneous users making API requests grow, I observe periodic TCPIP connection resets on a subset of API calls. In a typical HTTP application server, connection resources are finite, and backend processing can also cause holds on connection resources if there is synchronicity of an HTTP call to backend dependencies.
In any case, I believe that the proper response to an HTTP server's inability to handle more connections that are being requested is to issue 500 or 503 HTTP return codes, versus uncleanly dropping connections.