Thread by @PhDniX

ifiɣr pl. ifaɣriwn is one of those words that has unexpected i~a alternation in the stem between the singular and the plural. Compare also igidr pl. igadrn 'eagle', iɣirdm pl. iɣardmiwn. Such alternations also show up in Tuareg tenere pl. tinariwen 'desert'

https://twitter.com/JamalBahmad/status/1390454153252675585

Kb. izimr pl. izamarən 'lamb'
Tuareg teɣse taɣsiwen 'ewe'
Tuareg eskăr pl. askarăn 'nail'

Whenever such alternations show up, Tuareg consistently gives a reflex with /e/ in the singular and /a/ in the plural. /a/ and /e/ seem to be phonetically conditioned variants elsewhere.

Because there is otherwise no reason to assume historical apophony here, I argue that here too the /e/, that becomes /i/ in most northern varieties must originally come from an *a, that shifts to /e/ in the singular, while this is blocked in the plural.

There are also some apophonic plurals where the stem-vowel /e/ is treated as if it were *u. And important apophonic plural inserts *a before the last root consonant, and shifts every preceding *a to u. Some of the plurals of this type shift *e to u, e.g. Kb. idikəl pl. idukal

The rules of the conditioned *a > e shift are a little bit complex, but simply put: *a becomes /e/ if no other vowel occurs in the word except for *ă. Because all these plurals include vowels like *i or *a, the shift ends up blocked and *a resurfaces.

If you want to read the details of this shift make sure to check out my articles on the topic:
(this one only discusses the prefixes a- and e-)
academia.edu/10055169/_The_…
(this generalizes the shift, including plurals of this type)
academia.edu/38186227/Proto…

Until today, it's one of the articles that I'm honestly most proud of. It was incredibly hard to tease out an elegantly formulated version of what was clearly a sound law. It's still difficult, but it's really cool!