Ah. I've seen that in a bunch of places in Homer, but I didn't know it happened in Latin poetry as well.
It may not be quite the same thing in Homer. For example, in Il. 1.1
μῆνιν ἄειδε θεὰ Πηληϊάδεω Ἀχιλῆος
- u u - u u - - - u u ? u u - -
It looks like the εω should be scanned -, but this does not necessarily arise from a natural pronunciation of εω as a single syllable. Rather, at an earlier stage in the language Πηληϊάδεω would have been *Πηληϊάδᾱο. In this form the line would scan perfectly well, with the ο eliding:
*μῆνιν ἄειδε θεὰ Πηληϊάδᾱ' Ἀχιλῆος
- u u - u u - - - u u - u u - -
Early Greek hexameter tradition preserves formulae even when, as in this case, subsequent sound changes mean that they no longer scan properly. Of course, it's possible that later bards or rhapsodes reciting the line did pronounce it as a single syllable, but this would then be an artificial pronunciation created to explain the apparent metrical irregularity.
In the case of Vergil, there's no question of preserving earlier formulas from a preceding state of the language, so I'm guessing that there was a genuine Latin pronunciation of
dehinc as *
denc, or similar. Of course, it could also be an artificial pronunciation for the sake of the meter...