Cicero speaks of this as justified by the ear and by custom, rather than by reason:
[Cic. Or. XLVIII.] Quid vero hoc elegantius, quod non fit natura, sed quodam institute? indoctus dicimus brevi prima littera, insanis producta: inhumanus brevi, infelix longa: et, ne multis, quibus in verbis eae primae litterae sunt quae in sapiente atque felice, producte dicitur; in ceteris omnibus breviter: itemque composuit, consuevit, concrepit, confecit. Consule veritatem, reprehendet; refer ad aures, probabunt. Quaere, cur? Ita se dicent juvari. Voluptati autem aurium morigerari debet oratio.
In Donatus we have the same fact stated, with the same reason:
[Keil. v. IV. p. 442.] Quod magis aurium indicio quam artis ratione colligimus.
Thus we find numeral adverbs and others ending either in iens or ies, as centiens or centies, decies or deciens, millies or milliens, quotiens or quoties, totiens or toties. Other words, in like manner, participles and nouns, are written either with or without the n before s, as contunsum or contusum, obtunsus or obtusus, thesaurus or thensaurus (the ens is regularly represented in Greek by ης); infans or infas, frons or fros. In late Latin the n was frequently dropped in participle endings.
Donatus says that this nasal sound of n should be strenuously observed:
[Keil. v. IV. p. 442.] Illud vehementissime observare debemus, ut con et in quotiensque post se habent s vel f litteram, videamus quemadmodum pronuntientur. Plerumque enim non observantes in barbarismos incurrimus.
Gn in the terminations gnus, gna, gnum, has, according to Priscian, the power to lengthen the penultimate vowel.
[Prisc. I.] Gnus quoque, vel gna, vel gnum, terminantia, longam habent vocalem penultimam; ut a regno, regnum; a sto, stagnum; a bene, benignus; a male, malignus; ab abiete, abiegnus; privignus; Pelignus.
(Perhaps the liquid sound, as in cañon.)