Allen's Vox Latina, 2nd edition (1988) metions that there is occasional "poetic interchange" in Latin of syllabic [u] and non-syllabic [w], mentioning trisyllabic silua and disyllabic genva as examples ("Consonants", u, page 41).
I'd like to know more about where this interchange can be seen.
I found the following list of examples of -u- in place of -v- in An elementary manual of Latin prosody by William Ramsay (1859):
Lucretius has su̇ėmus, su̇ėti, su̇ėrit, su̇ėrint, su̇ėsse, su̇ȧdent, reliqu̇ȧs, reliqu̇ȯ, although ua and ue in these words is almost uniformly a single syllable in other poets. Catullus has solu̇it, solu̇u̇nt, dissolu̇ȯ, evolu̇ȧm, pervolu̇ėnt.
Tibullus, dissolu̇ėnda, dissolu̇isse, solu̇isse.
Horace, silu̇aė, su̇ėtae, milu̇ȯ.
Ovid, dissolu̇ȧntur, evolu̇isse, involu̇isse, exsolu̇isse, persolu̇ėre, persolu̇ėnda, milu̇u̇s. Lucan and Silius, Su̇ėvos, &c.
Some prosodians class with these relangu̇it [...] But this is a bad example, for in relanguit, the perfect tense of relanguesco, the vowels u and i always form separate syllables.
-qu- and -gu-
The most surprising part to me of Ramsay's list was reliqu̇ȧs, reliqu̇ȯ. I looked through De Rerum Natura in the PHI Latin texts corpus and I think the following lines are what the above quote refers to:
numquam relicuo reparari tempore posset. - De Rerum Natura 1.560
corpore relicuo pugnam caedesque petessit, - De Rerum Natura 3.648
relicuas tamen esse vias in mente patentis, - De Rerum Natura 4.976
In apparent conflict with these examples, though, a comment by Alex B. mentions a passage in The Phonology of Classical Latin (2020), where András Cser writes that syllabification of [w] to [u] "never happens to the [w] involved in〈qu〉or〈gu〉: a word like aqua or inguen is always two syllables" (page 31).
Is there possibly a difference between the behavior of〈qu〉or〈gu〉that come from PIE simple labiovelar stops, versus 〈qu〉or〈gu〉that come from a PIE labiovelar or velar stop followed by an originally separate [u] or [w]?
I found a reference in Adam's Latin Grammar to a trisyllabic pronunciation even of aquae, which seems truly incredible to me. The section on diaeresis mentions this line:
Quae calidum faciunt aqüae tactum atque vaporem. Lucr.
(Adam's Latin Grammar, by Allen Fisk, 1822, "Figures in Scanning", 4, page 186)
However, I did find sources that indicate that the reading aqüae is debatable, as manuscripts apparently have laticis instead of aquae and there is an alternative scan (discussed by Bede) of "aquae" as a disyllable with a heavy first syllable (1, 2, 3).
But then after finding those sources, I found a source that says that trisyllabic aquae is in fact found in other places in Lucretius, in De Rerum Natura 6.552 and 6.1072 (The Early Textual History of Lucretius' De Rerum Natura, by David Butterfield, 2013, page 81).
Fit quoque, ubi in magnas aquae vastasque lucunas - De Rerum Natura 6.552
vitigeni latices aquai fontibus audent - De Rerum Natura 6.1072
It looks like to make meter, aquai in this second line would have to scan as either trisyllabic *āquāī or quadrisyllabic *ăqŭāī; please let me know if I understood this correctly.
Aside from the specific matter of -qu-/-gu-, I would also be interested in any other information that leads me to a more complete understanding of this phenomenon (or these phenomena) and any applicable limits. A related past question by Joonas Ilmavirta♦, Synizesis in perfect tense 'ui', indicates that maybe we do not see -ui- > -vi- in perfect forms of verbs.
The Latin language; an historical account of Latin sounds, stems and flexions, by Wallace Martin Lindsay, 1894, page 87
"The Christianisation of Latin Metre A Study of Bede’s De arte metrica", by Seppo Heikkinen, page 128
De rerum natura, Book 6, commentary by Carlo Giussani, 1898