The main reason is speed. It ads extra latency that kills ad response rates.
Secondly is trust. The third party has no way to verify the first party isn't sending bogus requests to increase their payments from the third party and/or hurt their competitors.
Ad fraud is real. It goes the other way too, the "proxy" site could serve up a different ad from the one it's supposed to be sending, and it would be hard to know.
It only works if you have identified (logged-in) traffic and pass the identifier (ie, email hash) along to the side-channel.
Otherwise there's no way to target the ads, it's a first-party cookie but scoped to a single domain so there's no useful targeting data.
(and yes, this is the industry plan, but it's TBD whether publishers like media websites can get a workable fraction of their users to log themselves in)
