hi, Herodian’s (or pseudo-Herodian’s) De prosodia catholica (which is one of the key sources for our knowledge of grk accents) states expressly that ὦ is circumflex, and the e.g. he uses is before a vowel-initial vocative, ie. exactly the case you are considering:
“καὶ τὸ ὦ κλητικὸν περισπᾶται· “ὦ ἄνθρωπε”.”
A question I still have though is how to pronounce this in prose. Dionysius of halicarnassus treated ὦ before ἄνδρες as a separate long syll (see (i) below). In attic comedy however, you often see ὦ being blended into the next vowel-initial word, and commentators on attic prose texts have said that the same probably applies in prose (see (ii) below).
(i) Dionysius of halicarnassus takes this sentence from the opening of demosthenes’ speech on the crown:
“πρῶτον μέν, ὦ ἄνδρες Ἀθηναῖοι, τοῖς θεοῖς εὔχομαι πᾶσι καὶ πάσαις.”
He scans it as follows (i will explain in square brackets the metrical feet he refers to):
“ἄρχει δὲ τοῦ κώλου βακχεῖος ῥυθμός [i.e. long-long-short for πρῶτον μέν],
ἔπειθ’ ἕπεται σπονδεῖος [i.e. long-long for ὦ ἄν-: this shows that he treats ὦ as a separate long syll],
εἶτ’ ἀνάπαιστός τε [i.e. short-short-long for -δρες Ἀθη-],
καὶ μετὰ τοῦτον ἕτερος σπονδεῖος [i.e. long-long for -ναῖοι],
εἶθ’ ἑξῆς κρητικοὶ τρεῖς [i.e. three lots of long-short-long for τοῖς θεοῖς εὔχομαι πᾶσι καὶ],
σπονδεῖος δ’ ὁ τελευταῖος [i.e. long-long for πάσαις].”
(ii) on the other hand, in attic comedy you often see ὦ blending into the next syll. e.g. in aristophanes’ frogs you see:
line 58 μὴ σκῶπτέ μ᾽ ὦδέλφ᾽: οὐ γὰρ ἀλλ᾽ ἔχω κακῶς:
line 60 ποῖός τις ὦδελφίδιον;
line 164 καὶ χαῖρε πόλλ᾽ ὦδελφέ.
the acutes in lines 58 and 164 are bizarre; these are from the 2007 OCT, and there is nothing on the accents in the apparatus explaining this, however the accompanying text Aristophanea might discuss the accents here (i don’t have it). In any case these e.g.s show the blending of ὦ into the next vowel-initial word.
A more common e.g. of ὦ blending into the next vowel-initial word is ὦ ἄνδρες (into ὦνδρες), e.g. line 283 of aristophanes’ plutus:
”ἀλλ᾽ οὐκέτ᾽ ἂν κρύψαιμι. τὸν Πλοῦτον γὰρ ὦνδρες ἥκει”.
the scholion for this line (by tzetzes) says:
“ὦνδρες· ἀττικὴ συναλοιφή”.
As for the accent of the blended syllable, there is a scholion on lycophron noting specifically that it is circumflex in attic:
“εἰ δέ ἐστιν ἀττικὴ συναλοιφὴ ἀντὶ τοῦ ὦ ἄναξ, περισπᾶται διὰ τὸ ω τῆς κλητικὴς”.
As to whether this blending applies outside of poetry, the commentary of Dyer and Seymour on the apology (available on textkit, see pg 37, footnote 1) states that ὦ ἄνδρες was pronounced ὦνδρες. Cecil Wooten takes the same view in his 2008 commentary on demosthenes’ philippic I; on pg 41 discussing the first sentence of the speech, he says:
“Moreover, in this sentence, except for the frozen phrase ὦ ἄνδρες Ἀθηναῖοι, there is no hiatus, or gaps caused by the meeting of two vowels, to slow down the delivery, and there may not be a hiatus even there, since spoken Attic probably did a synizesis of the two vowels (cf. Smyth ss60-61)”. For the smyth reference, see this link:
http://www.ccel.org/s/smyth/grammar/html/smyth_1c_uni.htm#p21
nevertheless the question of whether or not to blend ὦ in prose is still unclear to me, as i don’t trust the modern sources if they don’t have ancient backing, and the ancient sources I have seen so far are not definitive on this point.
cheers 