This seems to be the same problem that many open-source projects have, where the md5 hash to verify your download is at a single, (often the same) location. One possible solution would be to use Twitter pinned tweet to also announce the address, however it's questionable how many people would actually cross check.
I always found that fascinating myself, serving the download and checksum from the same source. Doesn't http have enough redundancy checks built in to make this pointless? The only time to really do a checksum isn't on a browser download, it's when you push it over some serial connection, or android adb or something else.
But you're right about serving the data & checksum from the same source. I don't see what extra layer of security or integrity it adds. Someone tampering with server file system, or the data transfer (MitM) inherently has the access they need to inject their own MD5 into the HTML.