As the post points out, this only works if the site's content doesn't change very often, but for static sites this might be worthwhile.
Personally, I think that static sites are very easy to make secure, and sites that don't update often tend not to have much traffic, therefore don't make much money, and as such may not want to pay for this extra service (though I admit there are exceptions)
That said, if we could mark "expected change zones" and "expected stretch zones" in the images, such that blog posts or news articles would be exempt, then we could tell if it's a site hack or just new content... that might be usable. A second phase (and much research required) might be to detect hack-based content within these zones using some kind of text-based machine learning, comparing usual content and examples of the kinds of things hackers usually add.
Of course, this may be made harder in an American election year...