One concern I could see for browser based adblockers is the Server Push features of HTTP2, in which upon a client requesting a page, the server can deduce which subsequent requests a client will make (for example, if a client requests index.html, the server can assume that a request for styles.css and scripts.js will be coming shortly after) and 'push' them to client over the existing tcp stream without the client explicitly requesting them.
This could deprive an ablocker of the chance to deny content if I'm understanding it correctly. Then again, its very possible that I'm misunderstaning it. I should probably read the specification more thoroughly to answer some of my own questions.
Server operators are inclined to obey those hints, because otherwise they'd be wasting their own bandwith as well.
They gave an example of this with the "image" resource type: Whenever the engine loads an image (e.g., triggered by an <img src=...> declaration), those filters are triggered. It doesn't matter at all if the fetched resource actually has an image/... MIME type or not.
I could imagine the same thing working for HTTP2. The engine still knows which domain the request was initially for, even if all requests go to the same IP in the same connection in the end. Unless the filter says anything specific about the transport method, there is nothing stopping the engine from applying the existing filters to both HTTP and HTTP2. So domain-based blocking could actually become easier.