Many existing open source solutions are optimized for short lived HTTP requests and don't address the long running connection issue (like a large git clone). We wanted something better for our use case.