DOM in relative clauses

It is well know that Spanish exhibits Differential Object Marking (DOM), that is, some objects (in general, [+specific] and [+animate] objects) are marked with a preposition:

(1) Juan besó a la mujer

In (1), the preposition is mandatory. However, this effect seems to disappear with objects that are relativized:

(2) La mujer que Juan besó

However, if the relative pronoun is “quien” (that is, “who”, which is marked [+human]) the preposition is mandatory:

(3) La mujer a quien Juan besó

Interesting, the preposition is mandatory too if the relative pronoun gets a determiner:

(4) La mujer a la que Juan besó

This is true even if the nominal is non human:

(5) La gata a la que alimentaste

Notice that in these cases the relative clause is still restrictive. In fact, inanimate nominals (that cannot receive preposition in full clauses) cannot be relativized by using a determiner (at least the relative clause is interpreted non-restrictively):

(6) * La casa la que construiste [restrictive]
(7) La casa, la que construiste [non restrictive]

Can we take advantage if this to explore the structure of relative clauses, in particular, can we test if the relativized object is raised or base-generated using these contrasts?

Chomsky 2005

If you are interested, this is the last manuscript by Chomsky:
On Phases (2005), from the Minimalist Reading Group at University College London.

Infinitivals in subject position

Spanish easily accepts infinitival clauses in subject positions:

(1) Bailar es peligroso
(2) Bailar en el jardín es peligroso
(3) Matar un hombre en el jardín es peligroso

But with certain verbs this constructions is banned from the subject position:

(4) *Haber un hombre en el jardín es peligroso
(5) *Estar un hombre en el jardín es peligroso
(6) *Llover es peligroso

Intuitively, it seems that verbs that do not have a subject position cannot be in subject position. But this correlation is rather odd, since there is no reason why it should be the case that a verb that does not have a subject---like "llover," for instance---cannot be itself a subject.

The situation is even more complicated if we take in consideration verbs like "estar." Under standard analyses, this verb takes an small clause (SC), and the nominal inside SC raises to become the subjet of "estar":

(7) Estar [SC [un hombre] [en el jardín] ]
(8) [Un hombre] está [SC [t [en el jardín] ] ]

So we could say that (4) is bad because [un hombre] is not allowed to raise to become subject of "estar" (which is after all a non-inflected verb there). But what about "llover"? Is there any way to explain (4-6) in a uniform matter?