Thread by @PhDniX

A subtle but striking point that shows that the modern Arabic dialects cannot derive from Classical Arabic is a phonetic development Classical Arabic undergoes that is absent in most (if not all) modern dialects an assimilation of *a to respectively *i/u in CaCCī/uC nouns.

It's easy to find examples of this kind:
CA sikkīr 'drunken' Egyptian sakkīr
CA ṣiddīq 'truthful' Eg. ṣaddīʾ
CA ḫinzīr 'wild boar', Eg. ḫanzīr
CA ʿifrīt 'demon', Eg. ʿafrīt
CA ṣundūq 'box', Eg. sandūʾ
CA ʿuṣfūr 'sparrow', Eg. ʿaṣfūr
CA gumhūr 'crowd' Eg. gamhūr

Sometimes, even when modern dialects lack good parallels (or if they loaned from CA) it can be found that CA has innovated with this sound law.
CA tilmīḏ 'student' < Aramaic talmīḏ
CA dustūr 'register' < Persian dastūr
CA ǧibrīl 'gabriel' < Hebrew gaḇrīʾēl

What is striking though, is that the sound law does not apply universally (as sound laws normally do). Despite the stem II verbal noun having the exact same shape as *talmīḏ, namely, tafʿīl; in Classical Arabic that never becomes tifʿīl although one would expect it would.

The same is true for the mafʿūl passive participles, which one might expect to shift to mufʿūl in the same way *ṣandūq > ṣundūq etc.

So here we see clearly two competing outcomes of the same stem shapes, how can we understand this? It has to do with how CA came to be formed.

Classical Arabic is not a natural language: It's an amalgam of all kinds of different sources of "eloquent" Arabic. The selection of which forms entered the language was not a rational process but one dependent on the sense of aesthetics of the Arab grammarians.

At the time, the Grammarians were probably exposed both to the original modern dialectal forms, and these CiCCīC and CuCCūC forms. We can see traces of both forms being available sometimes in traditional Arabic dictionaries listing both forms.

For CiCCīC this is quite rare, but we find ǧibrīl besides ǧabrīl in the reading traditions for example.
CuCCūC adjectives/nouns often attest both forms:
saffūd/suffūd 'roasting fork'
qaddūs/quddūs 'all-holy'
sabbūḥ/subbūḥ 'all-perfect'
kallūb/kullūb 'roasting hook'

Eventually preference was given to one of these competing forms. It seems possible that explicitly the CiCCīC and CuCCūC forms were preferred BECAUSE the majority of the dialects did not have that shape, a concerted effort of exoticizing the high classical register.

Some people asked some interesting questions. First one asks if it occurs in other dialects than Egyptian. A small thread can be found here with some examples.

https://twitter.com/PhDniX/status/1078766278393978880

And here my answer to the question whether the CaCCīC in Egyptian is not due to guttural consonants. The answer is no; although it's a reasonable thing to worry about in many dialects.

https://twitter.com/PhDniX/status/1078769869364056065