One of the seven canonical readers as established by Ibn Mujāhid was Abū ʕamr, the reciter of Basra. His reading tradition is probably the most unusual one among the seven, lacking the hamzah in more places than any other reading tradition and having an extensive assimilation across word boundaries, which to some authors has suggested that the original tradition had no case vowels at all.
Today I'd like to focus on one specific feature of Abū ʕamr's tradition, namely his treatment of the noun rusul'prophets' and those that are like it.
Ibn Mujāhid has the following to say about it:
واختلفوا في ضم السين وإسكانها من قوله: (وَرُسُلِهِ ۦ) و(رُسُلُنَا).
فقرأَ أَبو عمرو ما أُضيف إلى مكنى على حرفين مثل: (رُسْلُنَا) و(رُسْلُكُم) و(رُسْلُهم) بإسكان السين، وثقَّل ما عدا ذلك. وروى على بن نصر، عن هرون عن أبي عمرو: أنه خفف (على رُسْلِكَ) أيضاً. وقال على بن نصر: سمعت أبا عمرو يقرأ: (على رُسُلكَ) ثقيلة.
وقرأَ الباقون كل ما كان في القرآن من هذا الجنس بالتثقيل.
"And (the readers) different on the u after the s or its absence among the words rusuli-hī and rusulu-nā.
Abū ʕamr read this word, if two letters were attached to it (i.e. a pronominal suffix -kum, -hum/him, -nā) with no vowel after the s (i.e. ruslu-nā, ruslu-kum, ruslu-hum), and he made heavy (i.e. added a vowel to) those besides this (i.e. when followed by any suffix that doesn't consist of two letters).
And ʕaliyy bin Naṣr transmits on the authority of Hārūn (who in his turn transmits) on the authority of Abū ʕamr that he would lighten (i.e. remove the vowel) ʕālā rusli-ka also. And ʕaliyy bin Naṣr said: "I heard Abū ʕamr read "ʕalā rusuli-ka" heavily (i.e. with the vowel).
And all the other readers read this type of word with the vowel in the whole of the Quran."
In light of ʕaliyy b. Naṣr's inconsistent account, it seems fairly safe to assume that the general rule is that the dropping of the vowel in rusul occurs only when it is followed by "two letters". Translating this to modern linguistic terms, a syncope of the second high vowel takes place when it is followed by a heavy syllable.
This has a couple of implications: First, the 3sg.m. form -hu/-hi which normally becomes -hū/-hī when it follows a short vowel, is treated as a light syllable, i.e. as -hu/-hi. This might suggest that it was originally read as such, and that the syncope predates this lengthening.
Second, this syncope is very difficult to understand if we assume 'Classical Arabic' stress. For Classical Arabic, final heavy syllables are 'invisible' to stress assignment. Stress is simply assigned to the first non-final heavy syllable, and if there is no such thing, stress is assigned to the leftmost syllable, hence: rúsulu-ka, rúsulu-hu, rúsulu-kum, rúsulu-nā. Which the stress, however, in exactly the same position, it is difficult to explain why words with a heavy final syllable would behave differently from each other, as the phonetic environment seems identical in any way that would matter.
Instead, it seems attractive to assume that a different stress rule operated on the Arabic of Abū ʕamr. This is not really problematic, there are certainly other stress systems within modern Arabic, and there is very little positive evidence to assume the Classical Arabic stress was even original to Classical Arabic. If we assume that stress always falls on the third morae from the right, taking a heavy syllable as countain as two morae, and a light syllable as one, we end up with a stress that would explain quite nicely the syncope as we find it:
rusúlu-ka, rusúlu-hu, rusulú-kum, rusulú-nā.
Next a syncope rule could have applied, syncopating any pre-stress (high?) vowel as long as it does not create a word-initial CC cluster:
rusúlu-ka, rusúlu-hu, ruslú-kum, ruslú-nā
Could the "and he made heavy that which followed it" mean: and he stressed the syllable that followed the syncopated vowel? To my knowledge no known terminology of 'stress' exists among Classical Arabic grammarians, but I am unsure how else ṯaqqala/taṯqīl should be interpreted here.
What is striking about this syncope, is that it is almost the opposite of what we would find in the modern Arabic dialects, where instead we would see rúsl-ak but rusúl-kum. Of course, the modern dialects are under no obligation to retain the full scope of stress systems of the 7th/8th century, but it is nevertheless striking to see a system so different from the modern dialects.
Seeing this rule, one wonders, of course, what happens to other nouns with the shape rusul when followed by heavy syllables. In a natural language (which Abū ʕamr's reading tradition certainly is not) one would expect a phonetic rule that applies to one word, should apply to every other word with the same shape. There are not many other words that qualify for this in the Quran, the only other words are:
subula-nā (Q14:12; Q29:69) 'our ways'
nuzulu-hum (Q56:56) 'their accomodations'
Ibn Mujāhid's description gives us no reason to assume that the above rule could be generalized to these words as well. And this shows a typical feature of the reading traditions; Very often the reading traditions are filled with these one-off dialectal forms, which are not part of any generalizable 'system'. Especially when so few words are affected by a phonetic rule in a reading tradition, the rule is not generalized as a phonetic rule as such. This in part must have to do with a kind of default ideal of Classical Arabic (whose conceptualization is never made explicit) from which the reading traditions are subsequently thought to deviate; the result is a crazy amalgam of linguistic features that could not possibly have been part of any natural language.
Interestingly, in this case, if we move to Al-Dānī's later account of this same rule, he actually has a slightly different description:
ابو عمر (رسلنا) و(رسلكم) و(رسلهم) و(سُبلنا) اذا كان بعد اللام حرفان باسكان السين والباء حيث وقع والباقون بضمها.
"Abū ʕamr (reads the words) ruslu-nā, ruslu-kum, ruslu-hum and subla-nā, when there are two letters after the final root consonant with no vowel after the s or b (respectivaly) when it occurs; The rest has a vowel u there."
In other words, Al-Dānī's account does include subla-nā, and seems to give a slightly more generalized rule of this syncopation.
But nuzulu-hum is still ignored. Interestingly, the expected nuzlu-hum is an "irregular" reading (i.e. non-canonical) attributed to Abū ʕamr by Ibn Xālawayhi, who says:
نُزْلهم بالإسكان هارون عن أبي عمرو وعياش
"nuzlu-hum with no vowel (after the zāy), (transmitted by) Hārun on the Authority of Abū ʕamr, and (transmitted by) ʕayyāš."
In this case it is therefore actually very likely that for Abū ʕamr there was a generalized rule of syncope as described above, but because of its rare application, the generalizable rule was forgotten, and the Classical Arabic form was assumed unless explicitly said otherwise. So if an account said "Abū ʕamr pronounce rusulu-nā as ruslu-nā", the logical step that this also included subla-nā and nuzlu-hum was simply no longer taken, and instead the Classical forms were assumed.
This is only a single case study, but ideally, one would like to see a proper linguistic description of the reading traditions, trying to find the 'rules' and 'exceptions'. As can be seen here, traces of a rule can be found, even when its behaviour is technically irregular and isolated.