Now that thanks to EU laws and browser imposing restrictions about third-party cookies it's more difficult, the whole "serve ads from other domain" may not be that relevant anyway.
If you use a random wildcard subdomain... just serve them from the main website, what is the difference? On the other side with a proxy just route the AD requests to another server if it needs to be, of course you have to find a way to distinguish which requests are for AD and which are not, something you can do with some sort of signature in the filename, so that only the server can know which requests shall be handled locally and which one forwarded to the AD provider server.
How much of the efficiency of online advertising comes from the actual "art" of tracking users and their preferences to display "personalized" ads vs the "efficiencies" from firing/outsourcing your marketing, ad sales and creative workforce.
selling ad space was always a lot of work. algorithms do it cheaper and in general better.
next step is just to run a GoogleAds lib/proxy...
I'm not surprised there are people who prioritize the latter, especially for small sites where they may not have someone who fully understands the risks.
Edit: A combination of DNS CAA with an account identifier restriction in the record would prevent this. Then the advertiser would complain, and any ads served would have to be over plaintext, which would cause browser warnings about mixed content and allow MITM injection of (more) malicious content.
And yeah, I can trivially block stuff in uBO by using CSS rules for example, so that's still on the table.
Higher impressions? Higher integration cost? I guess I'm not sure what the confusion might be. Advertisers obviously want to ram their bullshit down as many eyesockets as they can find.
They could have become the dominant advertisers online too, and then no doubt they'd be just as nasty. But they lost that war multiple times, first to doubleclick-likes and then to social media.