I was browsing the late Keith Smalls' DM1 (digitalmushaf.bodleian.ox.ac.uk) when I ran into this interesting vocalisation of الرسل in quran.com/38/14 where the primary vocalisation marks it ar-rusla and the secondary one ar-rusula. Only the secondary reading is canonical today. 
The Arab grammarians report for the dialects of Najd that they drop the second vowel in words of the shape CuCuC. We see this process being incorporated sometimes into Quranic readings (e.g. Ibn Kaṯīr's quds for qudus), but it never happens for plural nouns (rusul, subul, kutub)
But when we look at the description of Sībawayh (IV, 116, hārūn edition) of this process it is clear that he considered it a regular process that should occur equally in nouns of the shape CuCuC as in plurals with this shape.
This manuscript proves that using the Najdi CuCC plurals was still considered acceptable in reciting the Quran in pre-canonical times (this manuscript is likely mid third/early fourth c. AH), at least for the word rusul > rusl. This practice has completely fallen out of use.
In modern lexicons like Hans Wehr, the CuCuC and CuCC options are still frequently listed for singular nouns of this type, but plurals are never listed this way. Classical Arabic and the canonical readings reflect this same norm, cf. the ʿun(u)q versus rusul.
ADDENDUM: Incidentally, looking further, this is not just a fluke of this manuscript. I found two other manuscripts that show the same pattern unambiguously. I am sure there's more to be found! I wonder if it can be found for other plurals such as kutub.
