In Russian, the following is possible:
(1) V kakoj Ivan priehal gorod? In which Ivan arrived town? "In which town did Ivan arrive?"
I am afraid this might be a rabbit hole so I kept myself from tinkering too much here, but I want to lay this out.
Right now, my grammar can extract the determiner from the NP gorod. The HEAD there is nouny and is therefore MOD < >. Because it is MOD < >, it cannot participate as an adjunct in a head-adjunct rule.
If I were to remove this constraint from nouns generally, I suppose I could then try to carefully make sure it is put in place by the head-spec rule and by the bare NP rule? Would that be sufficient?.. This way, if the NP was formed by the extracted determiner rule, the MOD can be non-empty?
Then the question is how to attach the adposition. Normally an adposition’s complement is a noun, not a determiner. Does it sound perhaps like the adposition, being the head of the adjunct modifying the verb, is extracted here separately? Sounds weird but I currently don’t see how else to do it? But to do even that, the adposition would need to be QUE-nonemtpy, in other words perhaps it would need to share its complement’s determiner’s QUE value or something like that? (I think determiner’s share their SPEC’s QUE value as it is, so, if I am not seeing that in my case, that could just be a bug).