> There are tons of programming tasks that require coding literacy but do not require even knowing what a linked list is
I think people are assuming that I force the candidate to do this in C or something and munge about with next pointers and sentinel nodes. They're free to do so if they'd like, but usually they have a pretty high degree of control over what language they use and in this problem, what data structure it starts in (including arrays). As mentioned upthread, that means the problem literally reduces to something like loops, iteration, conditionals, collections. These abstractions are fundamental enough that again, any reasonable definition of coding includes them. Physics PhDs writing simulations occasionally have to deal with more than one particle (or whatever), and if they have no clue how to "do X for every Y", they won't get very far.
This isn't just hypothetical for me: I've interviewed PhDs (including non-CS ones), and pretty much all of them blew past this first filter. I've worked with people whose only prior programming experience was during their math PhD, and they generally were pretty great at programming (and pretty horrible at engineering). The latter is easy to teach (this guy became one of my best engineers) but the former is much more fundamental, much harder to teach, and generally something you want at time of hire.