Based on the phonetics of Romance languages, many of which show glides matching the frontness or backness of the preceding vowel between vowels that were in hiatus in Latin (discussed in "The Development of Latin Hiatus Groups in the Romance Languages", by Gail Keith Meadows, 1948, an article that Draconis linked to in an answer to this Linguistics SE question: Etymologically, why is there a v in “Giovanni”) it seems likely that Latin speakers had a [w]-like gliding transition between /u/ and a following vowel in hiatus rather than a glottal stop transition. That said, I don't know of any clear evidence that absolutely rules out the use of glottal stops to separate vowels in Latin.
One other thing that is relevant is that in certain time periods, at least, Classical Latin thought to have had a distinction in quality as well as quantity between long /iː, uː/, realized as high [iː, uː], and and short /i, u/ realized as [ɪ, ʊ]. We have evidence for this in the usual Proto-Romance reflexes of Latin /iː, uː/ as *i, *u vs. the usual reflexes of Latin /i, u/ as *e, *o. I believe there is also some evidence from Roman descriptions of the sounds of Latin, and there is some possible historical support also in that /u/ and /i/ in Classical Latin are often the result of vowel-reduction sound changes. (There is also evidence for the mid vowels having distinct qualities, /e, o/ [ɛ, ɔ] vs. /eː, oː/ [eː, oː], past a certain point—sorry, I forget the timeline.)
There is speculation that metrically short vowels preceding other vowels were however realized with a quality closer to the quality of a long vowel.
Meadows in fact wrote article about this topic: "Hiatus and Vocalic Quality in Classical and Vulgar Latin". I believe it is also mentioned in Allen's Vox Latina.
So "parvus" may have been pronounced [ˈparwʊs], and "metuunt" may have been pronounced something like [ˈmɛtuʊnt], [ˈmɛtʊwʊnt] or [ˈmɛtuwʊnt].
As Alex B points out in the comments, we have evidence that in some varieties of speech vowels that were orginally in hiatus sometimes came to be separated by a glide that wasn't homorganic to the preceding vowel; Meadows mentions evidence for this phenomenon in Romance languages and also mentions other kinds of non-glide epenthetic consonants like r in "afiree" for "afiee" ("promised") in Old French or d in Lombard (example: füdess < CL fuissem). I don't know if people have identified any regular patterns for epenthesis of that sort in Classical Latin.