Not knowing anything whatsoever about the esoteric subject of Old Welsh/Celtic historical linguistics, I’ve tried to read up on it a bit and hope tagging might attract here the attention of someone who knows better than me.
Looking at the (Romano-British?) name Cunetus, which seems most authoritatively to be at the root of the Welsh name Cynuit, Cynwyd and perhaps Breton Conet, relevant facts seem to be:
1. Cunetus was stressed on the normal penultimate syllable Cu-‘ne-tus
2. Some stressed vowels, in some circumstances, diphthongised (as in Romance); and in Old Welsh the unchecked -e- would diphthongise to -ui- (cf Common Celtic *ɸlētos > *luïd > llywyd) – not quite sure about this.
3. The final syllable being unstable (as in Romance), the hypothetical development might be:
Cu-‘ne-tus > *Cu-‘nuito
4. Final syllables dropped (again as in Romance) so:
*Cu-‘nuito > *Cun-‘uit and perhaps at this point the first (unstressed) syllable weakened from the Latinate oo [ʊ] to Welsh y [ə]?
5. The stress then shifted forward, so that it was once more on the penultimate:
Cyn-‘uit > ‘Cyn-uit
6. I think final unvoiced vowels became voiced (unlike Romance):
‘Cyn-uit > ‘Cyn-uid, ‘Cyn-wyd
These seem to be standard developments as outlined in David Willis, Old and Middle Welsh (2008?).
The point about diphthongisation would explain Breton Conoit (Conet), Langunuit, Langenewit (Langunnett), Llangynwyd (Llangoned); and the (unsourced) form St Knuet, which sounds like *Cy-‘nuit.
I am wondering if, when the stress shift took place, the (then) unstressed final diphthong simplified into a single vowel? This would explain why Langunuit/Langenewit became Langunnett and why Llangynwyd is also found in the form Llangoned. But there seem to be two important points:
i) that the original stressed vowel, that -e-, survives in some form: LangunnETT, LlangynWYD, LangonnET (Brittany, and still the stressed vowel there), LlandigwinnET. There is still a distinct syllable between the N and the T, even when it loses the stress.
ii) Not much evidence here, but at the time of Domesday, 1086, the spelling appears to have still included the diphthong (DB Langenewit, Exon Langunuit).
I expect it is clear now where all this is leading. If not I will stumble on in the next post.