The server today uses a thread-per-connection model. This means if there are a ton of idle clients, we will keep a ton of idle threads around and the goal of the ticket is to identify whether that is a significant cost.