liradi

A while back over at Lameen's place, we discussed the mystery morpheme -ij-/-iyy- that is used in Maltese and Siwi Berber with the plural suffix -at/-iet. Lameen argued that in both cases, it seems to be employed chiefly with nouns whose form is atypical for the given language. For Maltese, this would definitely make sense, since many of Maltese nouns with plural ending in -ijiet are borrowings. In fact, of the 20 most frequent nouns of this type in the MLRS Corpus, only three are of Arabic/Semitic origin: xogħlijiet = "works", mistoqsijiet = "questions" and aħbarijiet = "news". Sounds straightforward enough, especially in the context of Maltese where there is a separate conjugation paradigm for borrowed verbs, thus the existence of a noun suffix used predominantly with borrowed nouns with their strange and unusual syllable structures and vowel patterns is not that surprising.
This, however, cannot be the full story. For one, the question of 'what is a typical noun form' is one that is not that easy to answer. Secondly, Maltese is notorious - well, at least among us melitists - for applying some creative broken plurals to borrowed nouns. Thus we get forma / forom, storja / stejjer, spiża / spejjeż ("cost"), rotta / rotot ("routes") and so on, even though, say, storja with its four consonants or spiża with its very un-Semitic initial consonant cluster are not exactly, um, typical. At the same time, Romance borrowings of the honest-to-El Semitic CCVC / CVCC type (see Lipiński 2001:216) like skop or post form their plurals by means of the suffix -ijiet. Obviously there are other factors here at play, like perhaps the age of the borrowing or even the place and manner of articulation of root consonants, which would all have to be taken into account if a more detailed explanation is to be provided.
And then there's the whole semantic aspect. As Lameen notes in a reply to my comment where I wondered why we get art / artijiet ("earth, land") instead of *arieti (< Ar. 'arāḍī with presumed imāla, depharyngealization and devoicing) or żmien / żminijiet (instead of something like *azmina):

... "earth", "time", and for that matter "mother" are all words that are very rarely pluralised, increasing the pressure to adopt some commoner plural type.

This makes perfect sense - so in the evolution of Maltese, the original *arieti or something similar fell into disuse and artijiet, formed most likely by analogy, took over. And as I was reminded this morning by Charles Briffa's new book Iż-Żmien fuq Sider Malta, we actually have evidence for this shift. Not just any evidence, mind you, the mother of all evidence: Il-Kantilena. If you're reading this and you don't know about Il-Kantilena, feel free to consult Wikipedia for more details (I recommend the French version which seems to offer the most comprehensive account, and this image for the actual text). Suffice it to say that it is the oldest literary text in Maltese composed by Pietru Caxaro and dates back to the late 15th century (terminus ante quem 1485). We find what we're looking for on line 18:

Transcription (Wettinger and Fsadni 1968):
haliex liradi ’al col xibir sura

Modern orthography:
Għaliex l-iradi għal kull xiber sura:

Gloss:  
because DEF-land.PL for every span [1] shape

English: 
for each (piece of land) has its own shape (features)

And here it is: it would seem that ca. 1470, the noun art still formed a broken plural. One might consider - especially in the light of the genre - the possibility that Caxaro deliberately chose an archaism for both effect and reasons of metrics, but as for the latter, the number of syllables is the same for l-artijiet and l-iradi. In any case, it seems Lameen's hypothesis is correct and by 1796, the publication of Vassalli's Lexicon Melitense-Latino-Italum, art only had the suffixed plural form, in Vassalli's spelling Ardijyt. 
By the way, if you're wondering why it's iradi and not my hypothetical *arieti, it's because I forgot to account for the emphatic ḍ which, at least in most cases, inhibits imāla.
Finally, if you want hear what Il-Kantilena might have sounded like try the video below. The performer, Dr. Martin Zammit, is an Arabist and it kinda shows - for example in verse 18 (1:32) where the first word has -ie-, Martin reads [halāš]. Nevertheless, I think it is a pretty good approximation.

[1] span = "the space between thumb and forefinger". Cachia (1994:89) glosses għal kull xiber as għandhom = "they have".