I don't necessarily disagree with the heart of your statement, but it's worth mentioning that systemd does not require one to use resolved. The user is still "allowed" to run unbound, dnsmasq, etc., and the preferred solution will be a question for distribution maintainers and users to answer.

I understand concerns about systemd's scope, but most of the "questionable" functionality is optional.

Small comfort. When a project has the incredibly poor judgment (and, frankly, hubris) to write a caching DNS resolver without even the most basic cache hardening protections, and then declare it "a pretty complete caching DNS and LLMNR stub resolver"[1] instead of just leaving it to people who know what they're doing, it calls into question the rest of the project. This is a major reason why so many of us are wary of systemd.

[1] http://lwn.net/Articles/609740/

Personally, I'd prefer that they concentrate on stuff that's properly within the scope of the project. This is way, way outside what the reasonable scope of the project ought to be.