Dies and the fifth declension

Question

In an answer to this question on Ζεύς Draconis mentions (quoting his own post on another site):

The accusative form of the root, *dyēm (cf AGrk Zēn), also survived in Latin, in the form *diēm > diem. Uniquely, the dy here didn't become *dj > j; some scholars suggest that it evolved in a different Italic language/dialect, which didn't have the palatalization rules, then got borrowed back into Latin. When it did come back to Latin, though, it was with a new meaning of "day" (instead of "sky/sky god"), and it was extrapolated into a whole paradigm on the stem di-, and forced into the fifth declension where it didn't really fit (di-ēs, di-ēī, etc).

The very last mention caught my attention: dies was forced into the fifth declension where it didn't really fit. What does this mean exactly?

More specifically:

1. What was the original declension of dies in Latin like? How does it compare to third and fifth declensions at the time?

2. What kinds of words were in the original fifth declension? How does their declension differ from that expected for dies? The whole declension cannot be modeled after dies if it is an outlier of some kind.

3. Do we know why dies ended up in the fifth declension? Do we know of a mechanism that would do so or are there forms that scream fifth declension?

This all is of course under the assumption that the statement by Draconis is reasonably accurate. If it is not, then arguing to the contrary would make a fine answer here.

Answer 1

If the word for "day" had developed perfectly regularly, we'd actually expect to see it in the fourth declension, as it comes from a PIE u-stem! A few traces of this hypothetical fourth-declension are attested, such as the locative diū (which survives Classically in diū "all day") and the nominative dius (which survives Classically in nunc dius > nūdius "it is now the day").

The early Italic declension looked something like this (somewhere in between Proto-Italic and Old Latin):

nom  *dious
gen  *diovos/diovis
dat  *diovej
acc  *diēm
abl   ???
voc  *diou
loc  *dieu

The genitive and dative should look familiar: the forms with a consonantal j instead of a vocalic i became Jōvis and Jōvī.

If this had continued to evolve regularly, under the same analogical pressures that regularized the other u-stems, it might have turned into something like this:

nom   dius
gen  *diūs
dat  *diuī
acc   diēm
abl  *diū
voc  *diū
loc   diū

The form diēm sticks out, and doesn't fit in with the fourth declension at all. This oddity goes back to Proto-Indo-European (Stang's Law), and in most other words it got smoothed out by analogy: we see bovem "cow" instead of *bōm, for example.

But for whatever reason, diēm had more staying power than dius did. The word got rebuilt around this one form as an "e-stem": early *diūs became diēs, early diūī became diēī, and so on. This wasn't a particularly clean or regular process: the genitive singular is attested as diēs (cf the Greek first declension), diēī, diēi, and dieī, for example, with no clear consensus reached until post-Classical times.

Meanwhile, the "fifth declension" was slowly forming by analogy. Rēs was originally an irregular word that didn't fit into any declension pattern—and in the Sabellic languages (Oscan, Umbrian, etc), it's the only "e-stem" that's definitely attested. But in Latin and Faliscan, other irregular nouns slowly gravitated to its paradigm.

The details of this part are unclear and disputed, but eventually a regular declension pattern appeared on the model of rēs, with the productive suffix -ia gaining a variant -iēs to make new fifth-declension nouns (cf luxuria~luxuriēs, aciēs, maciēs, etc). This part never happened in Sabellic, which implies it was a purely post-Italic development without a precedent in PIE.

And once that declension pattern arose, diēs, with its very similar -ēs ending, was forced into it: the genitive became -eī, the accusative became -em, the ablative became -ē. Older forms like dius and diū died out, or were reanalyzed as separate words: diū was rebracketed as di- plus a suffix -ū "during", for example, which led to noct-ū "during the night".

In the process, diēs did leave its mark on the declension, too: the genitive singular of the whole fifth declension vacillates between diēs's -ēī and res's -eī. I don't think one or the other ever became truly universal until vowel length vanished and made it irrelevant.