I think Google prefers to minimize the: "if this domain, and only this domain, do this special thing". Like you said, given that it would be absurd to remove a UI element based on some whim that it'd put them in compliance, I'm sure a major element of the plaintiff's litigation case hinged on this very feature.
They have many cases of this. They actively filter based on domain to comply with takedown notices. That’s the most public example of how they have a huge number of exceptions based on domain and file.
Of course, I’m sure they do this in an intelligent way rather than literally a bunch of one off kludges.
Maybe completely ignoring a domain is different than a implementing subroutine that ignores indexing a domain's images but not its other (text-based) content?
