I was struck by the spelling of al-ǧuzʾ as الجزؤ (or actually الحرو) rather than the now common الجزء. This appears to have been a fairly common spelling in earlier times. Also Leiden's Or. 298, the earliest dated paper manuscript in Arabic (252 AH) has it too! 
It is not uncommon for nouns with the shape CuCC to have a CuCuC biform. And in fact, this very word is read as ǧuzuʾan by the canonical transmitter Šuʿbah of ʿĀṣim. This spelling may very well represent a pronunciation al-ǧuzuʾ rather than al-ǧuzʾ (despite the rasm).
A much more interesting interpretation of this spelling however might be suggested by al-Farrāʾ. For the word al-difʾ, he suggests that it may also be written with wāw for the nominative (al-difʾu), ʾalif for the accusative (al-difʾa) and yāʾ for the genitive (al-difʾi).
He suggests that this is done if one drops the hamzah, in which case the case vowels go to the previous consonant. Normally if this happens, it should not lead to a long vowel. To Sībawayh such forms would be al-difu, al-difa, al-difi, but al-Farrāʾ seems to suggest lengthening:
Citing as examples:
هؤلاء نشو صدق
hāʾulāʾi našū ṣidq "these are truthful youths"
raʾaytu našā ṣidq "I saw truthful youths"
marartu bi-našī ṣidq "I passed by truthful youths"
Spelling it like this is considered "correct" (ṣawāban).
هؤلاء نشو صدق
hāʾulāʾi našū ṣidq "these are truthful youths"
raʾaytu našā ṣidq "I saw truthful youths"
marartu bi-našī ṣidq "I passed by truthful youths"
Spelling it like this is considered "correct" (ṣawāban).
He admits, though that it is better (ʾaǧwad) to leave out the lengthening, since if you drop the hamzah in yasʾalu 'he asks', it is more common to say yasalu, rather than yasālu., and masʾalah 'problem' > masalah is more common than masālah.
This is in line with Sībawayh.
This is in line with Sībawayh.
There's some more interesting things going on here though. The example sentences with našʾ are not quite comparable to spelling الدفء as الدفو, since našʾ is always indefinite in these example sentences, and the wāw, ʾalif and yāʾ in fact stand in the place of case vowel+tanwīn.
Note that these pronunciations are NON-pausal. That is, not the form as attested in the last word of a sentence. Al-Farrāʾ and Sībawayh are frequently very careful in their example sentences, to add a final word to a sentence, if they want to show off the non-pausal practice.
This is what the function of ṣidq in the three example sentences are (incidentally these are the typical schematic examples hāḏā/hāḏihī/hāʾulāʾi for nominative, raʾaytu for accusative, marartu bi-X for genitive).
Sībawayh discusses a similar phenomenon; but here it is explicitly pausal:
Al-ridʾu, al-ridʾi and al-ridʾa show up in pause are pronounced as ar-riduʾ (also ar-ridiʾ to avoid CiCuC sequence), ar-ridiʾ and ar-ridaʾ.
Al-ridʾu, al-ridʾi and al-ridʾa show up in pause are pronounced as ar-riduʾ (also ar-ridiʾ to avoid CiCuC sequence), ar-ridiʾ and ar-ridaʾ.
The pausal form when dropping the hamzah, though, may replace the vowel+hamz with a semi-vowel instead, yielding: al-waṯw < al-waṯʾu, al-waṯy < al-waṯʾi but since an alif cannot follow a silent consonant, al-waṯʾa becomes al-waṯā.
The system as described by Sībawayh is similar but not identical. To my knowledge, Sībawayh does not describe using long vowels for C+hamzah+case vowel word-internally at all.
