> 2. Privacy. Their threat model only addresses malicious peers changing content, not malicious peers trying to track users.
I'm narrowing the scope around 'user tracking' specifically to 3rd parties, because 1st party user tracking is a requirement in order to perform some necessary functions (Providing accounts & user sessions, for example).
In the current centralized CDN model (Cloudflare, AWS, Akamai, etc.), the 3rd party is made clear (those aforementioned companies), and their need to track users is documented on the provider's ToS. Likewise, reputational damage is accrued to the providers in question.
In their proposed CDN model, there is no clear 3rd party, and so it must be defensively assumed that all of the data sent for requests is recorded at all times (The FBI Tor exit node problem). As a consequence, the minimal amount of request data can only be sent to these providers (IP address, content hash). Tor-like routing could be implemented to further disperse the requests from one IP address to multiple addresses, but now the monitoring problem's increased to the routers as well, on top of the additional latency introduced for said privacy. Mandating said routing into the protocol would make such tracks useless, as the IP address collected will roughly appear as random noise in the optimistic case.
Personal opinion below
However, it is likely that no technical solution provided would resolutely satisfy the question being asked.