To my knowledge, the compounds of jaciō are the only words where this complication occurs. And in Imperial Latin, these words frequently scan with a light initial syllable, indicating loss of /j/ and resyllabification of the final consonant of the prefix, such as /a.bi.ki.oː/.
Scholars of Latin seem to differ somewhat in how they explain the heavy scansion that is normal in Republican Latin poetry. The explanation that you mention in your question (VC.V syllabification of the first two syllables after loss of /j/, such as /ab.i/) is one of three slightly different explanations that I've seen for the heavy scansion. Some sources instead treat this scansion as evidence for the presence of /j/ and a VC.CV syllabification at the time when the poems were composed.
Possibly, the spelling "i" stood for the sequence /ji/ in Republican Latin
According to Alex B.'s answer here, which cites Allen 1978 and Leumann 1977, the scansion of prefixed compounds of jaciō with a heavy first syllable originally corresponded to a pronunciation with /ji/ like /ab.ji/.
It's possible that the spelling ⟨abi⟩ stood for /ab.ji/ due to an orthographic avoidance of double letter ⟨i⟩, which seems to have played a role in the use of single ⟨i⟩ as a representation of phonologically geminate intervocalic /jj/ (discussed by Allen a page away).
Possibly, a spelling ⟨ie⟩ and pronunciation /je/ were used in Republican Latin, and replaced in later manuscripts with "i"
I found an interesting but possibly outdated paper by Charles Exon about the pronunciation and spelling of this group of words: "The Form and Prosody of the Compounds of IACIO in the Present Stem" (1904, Hermathena, pages 129-162). Exon argues, partly on the basis of an Index of historical spellings compiled by M. W. Mather, that in the time period where compounds of jaciō regularly scanned with a heavy first syllable, they were pronounced with /je/. The explanation given is that regular vowel reduction would produce a form with /i/, but /j/ was reintroduced by analogy with the unprefixed form, and this caused a form with /je/ to develop by a special process of dissimilation the same way as /wo/ developed in words like servos (nominative singular). (Exon's view is that the use of ⟨o⟩ instead of ⟨u⟩ after ⟨qu⟩ and consonantal ⟨u⟩ was not simply an orthographic phenomenon, but indicative of a phonological process.) Later, this /e/ was raised and the analogically introduced /j/ was lost.
There are only a few examples of the spelling ⟨ie⟩ in surviving inscriptions and documents, but Exon finds them significant as he considers retention of archaic orthographic usage the best explanation for these rarely found spellings with ⟨e⟩ (pages 161-162).
Possibly, VC.V syllabification existed at some point
"Notes on Glide Treatment in Latin Orthography and Phonology: -iciō, servus, aiō", by Kanehiro Nishimura (Historische Sprachforschung, 2011), does argue that a VC.V syllabification like /ab.i.ki.oː/ existed for some speakers after the deletion of the glide /j/, writing "*ab-i̯aciō > (*)ab-i̯[ə]ciō > **ab-i̯iciō" to a final form where i̯ is double struck through, indicating glide deletion. According to Nishimura, "When the original morpheme boundary is retained in speakers' grammar, this prefixed verb may be scanned as HLLH (i.e., /ab.i.ci.o/)" (page 198).
I find this account somewhat problematic however because in all other contexts, Latin seems to have regularly preferred to resyllabify a consonant across a morpheme boundary to repair an empty onset. If the heavy scansion remained even after the deletion of /j/ because of psychological recognition of the original morpheme boundary, why isn't heavy scansion of the first syllable ever seen in even the most transparent compounds of originally vowel-initial verbs, which contain equally apparent morpheme boundaries?
After the loss of /j/ in Imperial Latin, an artificial poetic pronunciation might have existed with a lengthened vowel in the first syllable
Imperial Latin speakers who normally used /j/-less pronunciations starting with a light open syllable, such as /a.bi.ki.oː/, may have used pronunciations with a lengthened vowel, such as /aː.bi.ki.oː/, when reading poetry where the word has to scan with a heavy first syllable. Aulus Gellius complains about hearing speakers use unetymologically long vowels [oː] and [uː] for the sake of the meter in the first syllables of words like obiciebat and subicit (Attic Nights 4.17.1).
