This isn't a conspiracy - the problem with HTTP/1.1 pipelining is that responses have to go out in the same order as the requests that generated them. So, if a slow request comes in that generates a small amount of data, followed by 10 fast requests that generate a lot of data, the server has to hold all those big, fast responses in memory until the slow one finishes. That's bad for utilization - and it makes it easier to DOS a server. With HTTP/2, the server can respond immediately to whichever request finishes first - which is much better for utilization of server resources.