To me, it isn't too strange that it doesn't happen -- a shrewd design thief would know to strip tracking tags and anything that identified the original owner, just like a shrewd software pirate. They probably didn't realize what Munchkin actually was and didn't bother testing to see if the site worked without it.
When I say, "can't be prevented" I do mean from any level. Look at what Munchkin does: it tracks anonymous leads who hit your website. (Whether or not you later associate w/known leads, you start tracking when the lead is still anonymous.) So it needs to be loaded on a regular public page view. Anything on such a page can be scraped, definitely by a concerted human thief (and a smart machine can do this kind of thing, too).
You mention something like GTM helping conceal the Munchkin ID -- sure, it could make it harder for a human to find because it wouldn't be in the original HTML markup but injected on-the-fly. But if I scrape your page including the GTM tag, I get everything that comes along with it. Sorry!
Ultimately the only way to stop a stolen Munchkin code from having an effect on your stats would be for Marketo to discard unknown domains. But that would mean every one of us who's accustomed to just plugging Munchkin into all our client/internal sites, regardless of SEO aliases, typo-catching alternatives, etc., would have more work. It probably should've been this way from the start, but at this point few of us are going to want additional housekeeping tasks to prevent a relatively rare case.