From what I’ve gathered, sci-hub crawls each time a new paper is requested by the user (or retrieves from the cache if it’s previously crawled). But how do publishers not detect the high traffic sci-hub generates? How big is the pool of IPs sci-hub uses?