Accusative + Subjective Infinitive seems to be grammatical
Longmans' Latin Course: part III. Elementary Latin Prose, by
W. Horton Spragge, says that a subjective infinitive takes an accusative subject, and gives an example using "esse":
That you are happy is agreeable to me
Te beātum esse mihi est gratum
("Te" is the subject, and "beatum" is the predicative complement.)
Another example where esse is said to be a subjective infinitive taking an accusative subject and complement:
constat Romanos fortissimos esse
it is known that the Romans are very brave
(KURZOVÁ, H. (1986). Accusativus cum infinitivo in the structural-typological approach. Listy Filologické / Folia Philologica, 109(1), 1–10. http://www.jstor.org/stable/23462570, page 3)
So it seems to me "Hominem esse me delectat" would be correct.
However, other people also seem to have been uncertain about this matter; brianpck linked in the comments to http://latindiscussion.com/forum/latin/to-be-yourself-is-not-a-crime.11748/ where the user Bitmap provided the following examples:
Cic. inv.II,171: Nam aliter dicere solemus: "Necesse est Casilinenses se dedere Hannibali"; aliter autem: "Necesse est Casilinum venire in Hannibalis potestatem".
(Example of accusative subject of a subjective infinitive)
Terence, Heaut. 666: non licet hominem esse [saepe ita ut volt, si res non sinit.]
(Example of accusative subject of esse as a subjective infinitive; I extended the quotation to add the end of the sentence using the Packard Humanities Institute Latin Texts)
Cic.TuscI,9: necesse est enim miseros esse eos qui centum milibus
annorum ante occiderunt, vel potius omnis, quicumque nati sunt.
(If I understand the grammar right, it's an example of an accusative subject (eos) and complement (miseros) with esse as a subjective infinitive; I might have switched the complement and subject)
I don't think Nominative + Subjective Infinitive is a thing
Another discussion on that site http://latindiscussion.com/forum/latin/the-subject-of-the-infinitive.11753/ mentions some exceptions to the general tendency for the subject of an infinitive to be in accusative. However, as far as I can tell, none of the exceptions seem to apply in this case.
And in fact, my current understanding is that in the structure generally referred to as the "nominativus cum infinitivo", the nominative noun does not actually form a constituent with the infinitive; rather, the infinitive plays the role of a complement (not subject) of some other verb in the clause, and the nominative noun (not the infinitive) is the syntactic subject of that other verb.
The first exception listed is "Hercules fortis esse vult/videtur/dicitur" with "Hercules" and "fortis" in the nominative. I would interpret this the same way as the corresponding English sentences "Hercules wants/seems/is said to be strong': i.e., the grammatical subject is really "Hercules" rather than the infinitive.
This interpretation seems to be supported by patterns of verb agreement. As brianpck points out, in "nominativus cum infinitivo" constructions the passive verb is not always third-person neuter, as we would expect if the infinitive were the subject of the clause. I haven't studied this in depth, but an online worksheet I found (Infinitives, a.c.i. & n.c.i., Latin MILC – W1, MT 2012 October 09, 2012, from Robin Meyer's pages on the University of Oxford website) gives an example that shows the passive verb agreeing with a plural nominative noun:
- Parentes adire ad filios prohibentur. (Cic.Ver.2.5,117) Parents were prohibited to see their children.
You can see that the finite verb in this sentence agrees with "parentes" rather than with "adire". I don't have any examples, but I would guess that the same principle applies to gender agreement in compound tenses.
Actually, it seems that an impersonal passive construction with the accusative + infinitive may also be an option: see the following example cited in "Subject licensing in infinitival clauses: The Case of the Latin AcI" (Karen-Margrethe Dahl Hovind, 2020):
(5) b. Traditum est etiam [Homerum caecum fuisse]
'It has been claimed that Homer was blind, too.'
(Cic. Tusc. 5.39.114)
However, Sean Gleason ("Personal versus impersonal passive in Latin infinitival clauses: Some diachronic considerations", 2016) suggests that the use of the accusative + infinitive in this context, which "is found only from Cicero in the first century BC onwards", is "a partial Graecism").
Applying this to "Hominem esse me delectat"
So if I understand this correctly, the nominative + infinitive + main verb structure would not work to express "To be a human pleases me". I think "homo esse me delectat", if it's grammatical at all, would mean something like "A man pleases me to be" (i.e. "For a man to be pleases me"), which is pretty much nonsense. To express the proper meaning, I think it's necessary to make the infinitive "esse" the subject of "delectat" rather than the complement, and that means "homo/hominem" should not be the subject of "delectat", just the predicative complement of "esse". And in this context, it seems that the case of a predicative complement is assigned by the accusativus cum infinitivo rule.
I would interpret the infinitive in sentences like "Hominem esse me delectat" as having an implicit accusative subject like "me," and "hominem" is just the complement to that. (Or possibly, if it makes it feel more logical, you could even consider the following explicit "me" in "Hominem esse me delectat" to be like the subject of the infinitive, or at least the referent/"postcedent" of the complement. You can't do that for your second example "Finnum esse est canere" though.)