as has roughly been said, the shift from initial -ii to contracted -i in nouns of the form -ius occurred around the inception of the Augustan period. Preceding this period, then, are the comedians, Lucretius (about whom i will say more below), Virgil, Horace and (archaising) Persius and Manilius.
Propertius and Ovid, however, (followed ardently by Lucan) used -ii, with understandable metrical advantages (e.g. changing a once cretic, and thus intractable, form to a choriamb). a distinction was kept, though, between nouns and proper nouns, the latter typically retaining their single -i, perhaps from contemporary spoken practice. if one looks at the inscriptional evidence, -ii is found from the end of Augustus’ rule and has proliferated greatly by the time of Nero. nonethless, as Ritschl points out (Opusc.II.779), -i is still occasionally found up to the 3rd cent. AD.
a further distinction can be made of the pre-Augustans: whereas -i was invariably found in nouns of the type -ius, adjectives of the same form (e.g. patrius, medius) exhibited the uncontracted form in -ii, which can be seen, say, in Virgil and Lucretius (thrice).
as for the two instances commonly imputed to Lucretius, let us look closer:
the three instances of the adjectival genitive in -ii can, owing to the above distinction, be ignored. two instances, however, of an -ius noun with gen. in -ii are commonly cited. the first is at 5.1006:
improba nauigii ratio tum caeca iacebat.
the sentence stands in full grammatical completeness in this one verse and adds little to the preceding passage. the reading of the Oblongus(O) and Quadratus(Q) - both ninth cent. and in Leiden - is as above, and naturally scholars have been conservative about excising the line. the greatest Lucretian scholar, however, Lachmann, removed the line as an interpolation, on the basis of its being flat, largely irrelevant and, if this line be ignored, an anachronistic form of the genitive. Bernays and Brieger (and, i think, Giussani) treat the line with caution, bracketing it as a likely interpolation. for me this is a perfectly sound removal. Merrill, that crazy character, regards the line as genuine, on the basis of the testimony of O and Q, and the fact that somebody has to initiate the process of writing -ii and -i. Bailey does not bracket it, but gives voice to the likelihood of its being a rather poor interpolation, perhaps originally an interlinear gloss of sorts. Munro, the greatest British Latin scholar of the 19th cent., ‘emended’ nauigii to naucleti, which is generally agreed as an awful attempt to remove the problem.
the other instance, if it can be so called, is at 6.743: (in the fourth line below)
principio, quod Auerna uocantur nomine, id ab re
inpositum est, quia sunt auibus contraria cunctis,
e regione ea quod loca cum uenere uolantes,
remigii oblitae pennarum uela remittunt
praecipitesque cadunt molli ceruice profusae
in terram, si forte ita fert natura locorum,
aut in aquam, si forte lacus substratus Auerni.
this is, however, not the reading of OQ (of course our oldest testimonies). they carry remigio oblitae, which though metrically fine admittedly cannot make great sense in Latin. Lachmann’s solution to the problem was to posit remigi, with the long final -i, suffering correption (i.e. shortening of a long syllable in hiatus), and therefore scanning as a dactyl. this is not that rare in Lucretius, as horrific as it is to an Augustan poet. for instance, in the drn we find quae amara (uu-u) at 2.404 and twice in qui etesiae esse (uu-uu-u) at 6.716. this is an ingenious solution, and MS remigi could easily have been altered to remigio by diplography from the following o-. it was Marullus, of the fifteenth century, who emended remigio to remigii, no doubt unaware of the metrical solicism it entailed.
as for V.A.3.702 immanisque Gela fluuii cognomine dicta, i will say more tomorrow.
~D
I appreciate the quick and informative replies to my Q.