If your access to files consists of a public index (or something like a torrent tracker), they can just crawl the index, fetch all the keys, fetch all the files

If your access to files consists of a public index (or something like a torrent tracker), they can just crawl the index, fetch all the keys, fetch all the files and check them. It would be a bandwidth hog compared to running recognition patterns on localhost, but it's not hard. And more to the point, they could put the onus on Mega to go do that. It would be easy to argue in court "if we can find the key, so can they, if we can check the file, so can they" and cast their actions as wilful negligence. It's not so much whack-a-mole, as "keep it quiet", then. Mega pretty much has to kill what they can find the keys to, but if you keep the key a secret and share it sparingly, that is no longer possible.