One of the reasons why Poetic Arabic is often viewed as a supertribal poetic register, is the fact that the language seems to be 'the same' across different poets. In my IQSA blogpost I already challenged this idea. It seems very well possible that much of the dialectal variation that was present have been leveled out in the 9th century to serve the purpose of creating the idea of a single "Arabiyyah" that connected all Arabs.
However, there are some things that allow us to examine these questions; And we do find that sometimes we are able to find some amount of dialectal variation by paying close attention to the rhyme of the poetry.
As I have shown in my paper on the Arabic triphthongs it is clear that the two different spellings of ā, e.g. بنى <bny> banā and دعا <dʕʔ> daʕā actually go back to an ancient contrast, still retained in Quranic Arabic. This is clear because in Quranic Arabic these two spellings cannot rhyme with each other, suggesting that they are two different sounds. This is further confirmed by several Quranic reading traditions that indeed pronounce these two "ā"s different. The first as /ē/ and the second as /ā/. They also evidently have different historical backgrounds.
In Classical Arabic, however, both of these sounds have merged, and their distinction is purely an orthographic archaism. If we assume that Classical Arabic was indeed based on Poetic Arabic, we might expect that this hypothetical supertribal poetic register to have also lost the orthographic distinction between /ē/ and /ā/. We would then hypothesize that rhymes that end ā within Poetic Arabic would allow forms of both spellings. However, it turns out that this is actually very rare. I have found only a few:
The Najdi poetess Al-Ḫansāʾ who allegedly lived from 575 AD - 645 AD has a fairly short poem which clearly freely rhymes the two spellings of /ā/ with each other. We must therefore conclude that in her version of Poetic Arabic, there was no difference between the two etymologically distinct vowels.
The Hijazi poet Kaʿb bin Zuhayr who, coincidentally, also died in 645 AD, clearly shows that he cannot rhyme the two vowels. For example, this poem, consistently rhymes every final vowel with the /ē/. Whereas this poem, actually alternates between lines between the /ā/ and /ē/ vowel in a ABAB rhyming scheme; something rare in Arabic poetry, but very clearly intentional in this poem, proving without a shadow of a doubt that these sounds were distinct.
This dialectal difference between a Najdi poet and Hijazi poet goes completely unrecognised to the Classical commentators; To them there is no difference between the two /ā/ vowels that are spelled differently, and therefore they consider both to have poems with an ā rhyme. Considering this lack of recognition, it is all the more striking that the lack of distinction for the Najdi poet, and presence of distinction in the Hijazi poet aligns exactly with what the Medieval Arab grammarians tell us: Hijazi Arabic had a distinction between these two, while Eastern/Tamimi Arabic did not.
From this it is clear that there is at least some amount of dialectism in Poetic Arabic. At the same time, the very first line of Kaʿb's poem already contains a feature that cannot be reconciled with what we normally associate with Hijazi Arabic. Hijazi Arabic is said to have lost the hamzah, something that appears to be confirmed by Quranic Arabic based on the Quranic Consonantal Text, which indeed seems to lack the hamzah. But hallā saʔal[ti], which makes up the first block of the Kāmil meter, can only be scanned as | _ _ U _ |, getting rid of the hamzah: hallā sāl[ti] would yield the unmetrical | _ _ _ |. So despite a dialectism in his poetry, we cannot say that this poem was composed in the Hijazi dialect; It is clearly Poetic Arabic, with a Hijazism shining through.
We might still perhaps see this however not so much as a 'supertribal' language, but rather one of register. Perhaps in different parts of Arabia, Arabic speakers would simply have an archaic poetic register of their own local dialect; And that simply because this register was archaic everywhere, the language ended up looking more similar to each other than the spoken vernacular at the time. If this were true, we would expect, known differences in grammatical forms across different poems to not co-occur with the same poet and certainly not within the same poem. But we do find this.
One of the typical isoglosses between the Eastern/Tamimi and Western/Hijazi dialects are their deictic systems. It is said that the Eastern dialects would have the form ḏāذا for 'this' and ذاك ḏāka for 'that', whereas Western dialects would have the forms so familiar to non-poetic Classical Arabic and the Quran: hāḏāهذا for 'this' and ذلك ḏālika for 'that'.
In poetry, however, we actually find both forms of the distal demonstrative occur within a single poem within two subsequent lines, in a poem by the Tamīmī poet ʾAws b. Ḥajar.
فَذاكَعَتادي في الحُروبِ إِذا اِلتَظَت وَأَردَفَ بَأسٌ مِن حُروبٍ وَأَعجَلا
وَذَلِكَمِن جَمعي وَبِاللَهِ نِلتُهُ وَإِن تَلقَني الأَعداءُ لا أُلقَ أَعزَلا
So if the dialectal identifications are correct, we must conclude that the poetry is indeed mixing and matching dialectal forms within a single poem. That these dialectal differences are not completely fictional appears to be confirmed by the fact that within the Quran, a text that I would argue was composed in Hijazi Arabic, ḏālika occurs a grant total of 427 times, whereas ḏāka occurs zero times; A very different distribution of the kinds of things that we find in the poetry.
Let us take yet another feature; The third person masculine pronoun which in Prose Classical Arabic is normally hum and humu before a Waṣl has a variant humū as well; Having humū seems to pair with a whole slew of other forms as well. The pronominal suffix also becomes -humū/himū and the 2nd person masculine pronoun becomes kumū, -kumū and in the suffix conjugation of the verb -tumū. In the Quranic reading traditions we find some of the readers that use these forms consistently (e.g. The Meccan ibn Kaṯīr and the Medinan ʾAbū Jaʿfar, but not the Medinan Nāfiʿ!) in the position whereas others use the short forms consistently. In poetry, however, these forms are used right through each other, at the metres convenience.
Taking, for example the post-preposition -hum suffix, we find examples where within a single poem both hum and humū are employed, e.g. this poem by Kaʿb b. Zuhayr
Who reads: لَو يَعْلَمُ الأحْياءُ عِلْمِيَ فيهُمُ
Which is in the al-Kāmil meter which consists of two shorts OR a long, followed by a long, short long 3x in a row:
_ _ u _ | _ _ u _ | u u _ u _ ||
law-yaʕ-li-mul-|ʔaḥ-yā-ʔu-ʕil-|mi-ya-fī-hu-mū||
Only two lines down we find: يَتَطَهَّرونَ كأنَّهُ نُسُكٌ لهُم
u u _ u _ | u u _ u _ | u u _ u _ ||
ya-ta-ṭah-ha-rū-|na-ka-ʔan-na-hū-|nu-su-kun-la-hum||
This free variation between the two forms does not seem to get described by the Arab grammarians who either attribute it to one dialect or the other; This free mixing is unusual, and for example, not what we find in Quranic Arabic nor the Arabic of the Psalm fragment. The Quranic reading traditions also either have -hum or they have -humū and never both. So clearly, whatever the poetry is doing, it is doing something different in this regard than prose from the early Islamic period.
===
So how should we understand these things? If we trust the Arab grammatical tradition, we must conclude that what they describe as 'dialectal features' can occur side-by-side in poems. There are however issues of authenticity here. By the time the Arabic grammarians were working, there can be little doubt that the "dialects" they are describing were already not in the form anymore that they envision them to be. It could therefore very well be that the dialectisms are not from actual knowledge or hearsay but extracted from the poetry of poets of these different regions. That is still to some extent meaningful, but it could mean that when the Arab grammarians are talking about dialectal features, they are actually talking about regional poetic licenses of the Poetic language.
It seems possible to envision that at first there was a single performance register; quite archaic, not dissimilar to Epic Greek. The oral formulae and the strong metrical structure caused the archaic structure to be preserved. However, as the spoken language developed further, new forms which were occasionally metrically convenient would come in existence in the spoken register, which were subsequently incorporated into new poetic formulae in the poetic register creating a poetic style which incorporates all kinds of archaic and innovative forms within a single poem. This need not mean that the text was post-hoc amalgamated with the collection of these poems, this could very well have been part of the poetic register of the poets that these poems are attributed to.
The question whether besides 'archaisms' you could also incorporate formulae from different dialects into your poetic register is something that seems generally well accepted to exist in Epic Greek, although those ideas are being challenged at the moment. But a much more careful look into the poetry and a careful cataloguing of the different 'dialectal' forms is extremely necessary. How often do these forms overlap? Certainly ḏālika is a dialectal form, but its probably also an innovative form compared to ḏāka. How often do we see non-Hijazis use ḏālika? Can we draw a "dialectal map" independent of the Arab grammarians and come to the same conclusions? How strongly mixed is it?
All questions for a later day (and probably a new research grant...) but it is clear that the poetry requires a much more rigorous linguistic analysis than it has so far been subjected to.