Thanks for your response! I see now that the constituency waters are murkier than I thought.
I believe, these tests fail also for #27 (1), where intuitively we do have a constituent (headed by an infinitive), don't they? Therefore, I don't think, they have the necessary power to adjudicate this case.
I would presume they would fail, but I can't really judge this. Even the unmodified #27(1) form of these obligationary -naa constructions are a bit "exotic" for me (as in, I've certainly run into them in both speech and writing, but I actively notice them every time I run into them, and I would never say these forms myself). But it sounds like you've had more experience with these forms, so I trust your judgments!
---
I also agree with you about the power of (pseudo-)clefting for HU constituency, at least in the form that I attempted to formulate it in #30. The "output" of this clefting operation seems to often be kind of awkward, even when what's being clefted is clearly a constituent, like "uskii naii kitaab" in the following:
(A) maiN uskii naii kitaab paRh rahaa huuN -> (???)uskii naii kitaab hai jo maiN paRh rahaa huuN
In other words, the clefting test as formulated in #30 seems to have a high "false negative" rate.
Maybe there's a better way to cleft. Or maybe not. I guess, in English, clefting is something that one would do to shift the topic-focus structure of a sentence out of neutral. In UH, scrambling is the typical way of accomplishing this goal, so maybe clefted sentences will always sound a bit off.
---
I feel fewer qualms about pro-form substitution, but it does lead to some interesting observations.
It seems like when we have something that's clearly an infintive constituent, it is substitutable with a pro-form like ye.
(B) kapRe dhonaa aasaan hai -> ye aasaan hai.
(C) merii qismat meN kapRe dhonaa likhaa thaa -> merii qismat meN ye likhaa thaa.
(D) mujhe kapRe dhonaa aataa hai -> mujhe ye aataa hai.
Now the thing I find interesting is that... To me, sentences (C-D) also allow variants where the -naa form with its object. But, once the -naa form agrees, pro-form substitution seems to get blocked!
(E) merii qismat meN kapRe dhone likhe the -> *merii qismat meN ye likhe the.
(F) mujhe kapRe dhone aate haiN -> *mujhe ye aate haiN.
The sentences on the right are okay abstractly (eg, someone might say "mujhe ye aate haiN" when they want to say that some set of clothes do in fact fit them), but they don't feel right to me if ye refers to the "kapRe dhone."
This seems to suggest that maybe the -naa forms in (E-F) are less constituent-y than the -naa forms in (B-D). I'm surprised by this, since my intuition was that the latter are actually infinitives (cf. #21). Murky waters...!
Actually, maybe this pro-form substitution is actually testing something quite similar to the "answer fragments" test. Afte all, "kyaa" is also a pro-form! I kind of suspect that the results of the pro-form substitution test, when applied to -naa clauses, might always agree with the result of the answer fragments test.
But you may be right that these judgments may not be universal. And in any case, this doesn't change the fact that maybe this doesn't give the right answer for the obligationary constructions from #27(1).
Actually, for me this sentence works as long as you add a "bhii" after aapko.
The bhii should definitely have been there. It's interesting that this sentence (copied as (H) below) works for you! It still sounds a bit odd to me. What about (I-J)...?
(H) mujhe khilaune khariidne paRenge aur aapko bhii paRenge.
(I) mujhe khilaune khariidne haiN aur aapko bhii haiN.
(J) mujhe khilaune khariidne chaahiye(N) aur aapko bhii chaahiye(N).
I expect there'd be no disagreement that all of these are acceptable without their final words (paRenge, haiN, chaahiye(N)).
---
To me, ellipsis again seems to work okay when we clearly have an infinitive constituent.
(K) kapRe dhonaa aasaan hai + kapRe dhonaa mazedaar hai -> kapRe dhonaa aasaan hai, aur mazedaar bhii hai.
(L) merii qismat meN kapRe dhonaa likhaa thaa + aapkii qismat meN kapRe dhonaa likhaa thaa -> merii qismat meN kapRe dhonaa likhaa thaa, aur aapkii qismat meN bhii likhaa thaa.
(M) mujhe kapRe dhonaa aataa hai + aapko kapRe dhonaa aataa hai -> mujhe kapRe dhonaa aataa hai, aur aapko bhii aataa hai.
Again, (L-M) allow variants where the -naa form can agree. But now, the resulting clauses *do* seem to act like constituents under the ellipsis test.
(N) merii qismat meN kapRe dhone likhe the + aapkii qismat meN kapRe dhone likhe the -> merii qismat meN kapRe dhone likhe the, aur aapkii qismat meN bhii likhe the.
(O) mujhe kapRe dhone aate haiN + aapko kapRe dhone aate haiN -> mujhe kapRe dhone aate haiN, aur aapko bhii aate haiN.
This makes it rather unclear to me whether "kapRe dhone" in "mujhe kapRe dhone aate haiN" is a constituent, since it fails the pro-form substitution test but passes the ellipsis test. Murky waters again...!
---
Anyway, in summary, you're right that constituency here is fuzzier than I made it out to be. Even for one person's idiolect (eg, mine), constituency tests sometimes don't give consistent results. And the tests sometimes output rather "corner-case"-ish test sentences, so different individuals' acceptability judgments will probably also vary.