3, 4 and 5 do not work, no. was für is a particularly devious construction. Even 2 strikes me as wrong, nevermind 1. That is, as for 2, it rather reminds of Was hat das heute viel geregnet (oh it has rained a lot today).
As for 3 etc, one should regularly use welch- to ask for a joined group
Welche Bücher werden von was für Menschen gelesen?
The opposite distribution too appeals prima faci
Was für Bücher werden von welchen Menschen gelesen?
There is no difference in practice, as far as I can tell. An empirical study with detailed surveys might show preferences, but these should blush in comparison to use of welche throughout.
The difference that I have in mind is incongruent. On the one hand, welche should introduce subclauses. On the other hand, it's a primary question word that should lead the question. These are simply two different usage patterns. There is no real subordination if both phrases are codependend. On the other hand, welch supposedly reflects the same root as Engl. like "similar", etc. This suggest the expectation of a certain community as coreference. Whereas was für asks for diversity. This parallels existential and universal quantification in logic. Existential and universal normal forms (leading with either) are however equivalent up to isomorphism, so that's no constraint on preference. Howsoever, this also roughly reminiscent of in-/definite determiners, welche Menschen Lesen _ Buch?, Menschen lesen welche Bücher?. Obviously we are drawn to assign agency primarily to humans as subjects. There's the Mantra that the most important part of a sentence should come last; conversely, the variable symbol that is in question, so far devoid of, yet to be filled with meaning comes first in a question.
Was für Bücher lesen Menschen.
Yet, the reverse does not work
* Welche Bücher lesen Menschen? No book has ever read a human.
This is less clear when the difference between agent and target does not exist. Compare Mario fand Silvia immer attraktiv, which interpretation depends on context, IMHO, in part on the sex of the reader.
Further, the wh- element in which ~ welch might just compare to the possibility-marking wohl (ja, das wird wohl so sein) and Dutch wellicht "vielleicht" (well likely, maybe) and even further to pronomial one, indeed ähnlich, alike
welche Menschen lesen was für ähnliche Bücher
and, if ähnlich compares to sel(b)-, then denselben comes to mind, but this would be misconstrued as backreference to a supposedly aforementioned selfsame group like people. Instead, sel- compares indirectly to En. same, Ger. sämtlich, or somanche, En. somany (though not ?some any), then so comes to mind
und, was ließt du so für Bücher? (*as for books, what do you read yourself?)
Was für Bücher lesen Menschen so?