Problems of the traditional explanation
CVCC form presupposes consonant clusters which are otherwise unattested in Hebrew. The Masoretes resyllabified the words with dagesh to break even a hint of consonantal clusters (ni-zchar - niz-car).
Consonantal clusters depend on non-aspirated variety of begedkefet (calb is pronounceable while calv is not). But we know from the LXX and Hexapla that begedkefet letters were mostly fricatives. (They lost aspiration after the Masoretes introduced dagesh kal.)
The CVCC hypothesis does not explain qatb/qitb/qutb forms but merely postulates their existence. It does not explain why a particular word has [i] while another word [a]. In Hebrew, vowels are always morphological (davar/diber/cotev/gadol, etc). It runs against the language’s logic that only in qatb/qitb/qutb vowels lack morphological importance: no semantic difference has been shown for qatb/qitb/qutb.
Hebrew vowels in CVCC words are not always the same as in Arabic (Hebrew ragl, Arabic rigl). The correlation of Hebrew and Akkadian CVCC vowels is not that strong and could be affected by the influence of Hebrew on reconstructing Akkadian vowels.
Reductions and dagesh
Except for CVCC, only long vowels (kamatz/tzere/holam) are reduced to schwa synchronically in the MT. CVCC is claimed as the only exception to that rule (sifri - sfarim, calb - clavim). This cannot be plausible also because shwa can expand to hirek (dvarim - divrei Torah), thus the opposite process of hirek reducing to shwa is unlikely.
The process which reduces a short vowel to schwa yet simultaneously introduces a long epenthetic vowel (sifr - sfarim) is otherwise unattested. Vowel assimilation is unattested in Hebrew on the level of morphological forms.
Long vowel in epenthesis would be unlikely to appear when its short counterpart, patah is readily available.
Reduction of epenthetic vowel (sfarim - sifrei) would be very strange: if the vowel had been so necessary as to cause epenthesis, it is not easy to see why it would have been lost.
The notion of shwa meraheph does not explain calb - clavim variation but merely clothes the problem in scientific terms.
Problems of strict phonological explanation
The choice of a/i/u in CVCC cannot be predicted based on the first two radicals (karn, but kirv). Neither it can be predicted based on the final two radicals (carm, but zirm). Combinations of begedkefet can also be similar between the two groups (calb, but pilc).
Somewhat more complicated explanations are also not easy to build. For example, in chalbh every consonantal sound originates higher in the vocal tract than the previous sound, and in phlbh the direction is opposite. But in malch the sound’s origin is descending, thus the hypothesis that a/i choice depends on sequence of consonants is problematic.
There is some evidence from polling singers that the Masoretic choice of vowels for CVCC is comfortable for chanting. Singers preferred calbi to cilbi, sifri to safri, etc. While potentially promising, this approach requires more extensive studies.
CVCC appeared from CaCaC with both kamatz vowels
Virtually no roots are present both in qitb and qatab forms, though the roots of qitb form are commonly present in other nouns forms (derech - modern madrich, but not darach). This suggests that most qatab nouns transformed into qitb. No semantic difference has been suggested to explain an opposite hypothesis: that some roots formed qitb-type nouns while others evolved into qatab nouns.
The direction of transformation is from qatab to qitb because qatab is an exceedingly ancient form harking back to the time when the language did not differentiate between nouns and adverbs (davar, rahav, halav, but also lavan, katan, halaq). The remaining qatab-type words which did not evolve in qitb/qatb might not be nouns after all, but participles.
Why a few nouns survived in qatab form is uncertain. An equally good question is why qatab-type adverbs did not evolve in qatob. Perhaps, qatab nouns were commonly employed in smihut form which inhibited word-initial stress shift.
Several nouns encountered both in qatab and qatb forms have very close meaning (davar - dever, halav - helev), which hints at their common origin. Indeed, davar is mostly employed in smihut while dever stands alone. The relation is less clear for halav - helev, though halav is commonly employed in possessive form and in “milk and honey” idiom with forward stress shift.
Originally, Hebrew had moraic bisyllabic stress with final syllable accented. In vulgar speech, word-initial stress shift in nouns moved the accent: calAv - cAlav.
In the absence of gemination (dagesh hazak), post-tonic vowel was reduced: (calAv - cAlav - cAl’v) - calb - calbi. The first stage is davar-type noun, then word-initial stress shift, then reduction of post-tonic vowel, then loss of vocal shwa as the final consonant evolved from fricative into a stop, then possessive forms became construed synchronically from calb form. (Compare -ck digraph, which denotes short preceding vowel in English, but long vowel in German.)
Plural suffix is always accented (-im accented as the final closed syllable; morphological tzere stressed generally in strong roots; -ei stressed additionally because of forward stress shift in smihut). No word-initial stress shift, therefore, had occurred in plural forms clavIm, calvEi. Absence of a consonantal cluster (which is present in calb) accounts for the absence of dagesh in clavim, calvei; no need for the concept of shwa meraheph. To put it differently, the singular form developed as malAch - mAlach - mal’ch - malc (with a dagesh after silent shwa). The plural form never suffered word-initial stress shift: malAch - m’lachIm - m’l’chEi israel - mal’chEi israel (no need for dagesh).
Reduction is made simple: just as we would expect, first kamatz shortens in plural, second kamatz in smihut.
Second vowel and dagesh in segolates
Secunda lacks the second vowel in any segolate nouns. The segol seems to have been introduced by the Masoretes to facilitate chanting.
Neither does Secunda support the Masoretic dagesh: derxo instead of darco, Melxa instead of Malca. The Masoretes mistook the schwa for silent, and therefore introduced dagesh. Absence of dagesh in Secunda’s segolates supports vocal schwa: calAv - cAlav - cal’v.
Vowel variation: patah in the words of relative importance
Secunda shows no correlation in CVCC vowel with the Masoretes: sams instead of sims, bekr instead of bukr, derx instead of darc, and ars similarly to aretz. The vowel in CVCC words thus remained indistinctive after some 1600 years of Hebrew linguistic history, and it is unlikely that that vowel would have been fixed for each word by the time of the Masoretes while the language was no longer spoken. The Masoretic a/i/u distinction in CVCC nouns has, therefore, to be taken with a grain of salt.
The Secunda’s vocalization disproves a hypothesis which traces qatb/qitb/qutb to proto-Semitic forms. Such differentiation did not exist even in the fourth century.
As can be seen from Secunda, the vocal sound in CVCC was not a clear [a], but rather an indistinctive segol- (or ы) type very short [ae]. It would take only a small difference in pronunciation for the Masoretes to prefer [a] to the standard [i] in a few words. Yemeni Hebrew and Babylonian notation do not differentiate between patah, segol, and vocal shwa, which also suggests that the sounds are very close.
Smihut, possessive, and feminine forms all employ the same vowel (malcei - malci - malca). This cannot be a coincidence. The Masoretes heard [a] in one form and transplanted it to the others. The donor form could not have been feminine because assimilation would have led the Masoretes to hear [a] in many more words (kvasa would sound more natural than kvisa). Possessive form is relatively rare, unattested for a number of CVCC words with [a]. Thus, the donor form was smihut, a very common form.
Normally, smihut has intonational stress on final word (this explains stress shift forward and reduction, dvarim - dvrei Torah - divrei Torah). But not so when the first word is very important: in melech haaretz, “king” is pronounced stronger than “land”. It follows that smihut stress shift forward does not leave words of importance completely unaccented.
If the word melech were intonationally neutral, it would have been reduced like this: mlachim - mlchei haaretz - milchei haaretz. Since the word “kings” retains intonational stress, the first of two schwas is expanded somewhat more than usual. Normally, that shwa sounds like a very short [ae], represented by hirek; the Masoretes recorded a bit longer [ae] as patah.
The expansion of shwa to patah (clvei - calvei) is not without a precedent: intonational stress of the recital expanded shwa under waw in wayomer to patah.
Tzere normally shortens to patah in closed accented syllable (diber - dibArta). Forward stress shift in smihut left most segholates unaccented, and tzere shortens to hirek in closed unaccented syllable. A few segholates which are emphatically stressed in smihut retained stress on their closed first syllable, and have patah instead of hirek.
A slightly esoteric concept of sound symbolism in phonaesthesia also recognizes [a] as more important and respectful sound than [i].
The relative importance can be defined as the word being more semantically important that the following word in smihut. Thus, in regel yosef the semantical accent would fall on “foot”: it will be clear from the context that the discourse is about Yosef. The smihut’s “news” are about the foot. The test of relative importance is this: such word cannot be excised from the text without impairing its meaning, while the following word can be safely omitted (regel yosef ~ raglo).
The semantic peculiarity explains why very few Hebrew words have qatb rather than qitb form. Semantics of a typical smihut might be different in biblical narrative (where the Masoretes heard it) and speech. That may account for such variations as Hebrew ragl vs Arab rigl: in Arabic, a spoken language, a mention of rigl abdullah might not have been preceded by a reference to Abdullah. A smihut, thus, would have an emphatic stress on the name rather than “foot.”
Semantic peculiarities of certain words in smihut could have produced the observable correspondence of vowels between Hebrew and Arabic qitb/qatb words. Only the words which typically retain intonational stress in smihut have [a] in CVCC.
First vowel in segolates
The Masoretes could only hear the first vowel as tzere if the consonantal cluster was weak: sef’r, but sifr; very long tzere could not survive in heavy CVCC syllable. They recorded the shwa-like sound as segol because shwa was reserved for reductions.
The original kamatz evolved into tzere through elongation in open stressed syllable: safAr - sAfar - sa:f’r - sef’r.
Vocalization is different in segolates of respect (sifr but malc) because second segol there is not a shwa-like sound but fully epenthetical. The Masoretes resorted to epenthesis to make the segolates of respect easily chantable on the model of regular segolates. The epenthesis occurred at the expense of reducing the first vowel. Thus in segolates of respect the first vowel, tzere was reduced to segol. Tzere is normally reduced to patah when not to shwa (diber - dibarta). There may be two reasons for the difference. One, epenthetic reduction might change tzere into segol while moving stress reduction changes tzere into patah. Two, the Masoretes wanted the segolates of respect to sound not dissimilar to normal segolates, thus reduced tzere to segol to preserve the CeCeC pattern.
The absence of vocal shwa in segolates of respect (malc but sif’r) was due to the chanting tradition. By analogy with smihut, the Masoretes imagined or heard slightly forceful pronunciation of segolates of respect. In such forceful pronunciation, vocal shwa would be squeezed out and become silent.