A little while back, I asked a question about the alleged Latin word "tribalis" (which it seems was not actually used), and I mentioned that it seemed to me that it would be an irregular formation because tribus is a u-stem noun, but tribālis does not contain -u-. So far, nobody has informed me of any examples of parallel formations of that type.
But recently, I did run into an example of the reverse: an o-stem noun that has a derivative that seems to be built with -u-. It is lectŭārĭus, which Lewis & Short says is a "late Lat." adjective meaning "of or belonging to the bed, bed-".
The related noun lectus seems to be declined as an o-stem. That said, I found some interesting notes in the L&S entry that may be relevant: it says
(nom. lectum , i, n., Dig. 32, 1, 52, § 9; 34, 2, 19, § 8; lectus , ūs, Plaut. Am. 1, 3, 15; al. lecti; Sen. Ep. 95, 72 Haas; Cornif. ap. Prisc. 711 P.)
The relevant line seems to be
prius abis quam lectus ubi cubuisti concaluit locus.
(Amphitruo (Leo), Wikisource)
The "lectus" here apparently should be read as a a genitive.
But I don't know if usage in Plautus would be relevant to a "late Latin" derived adjective.
Oddly, de Vaan's entry for lectus doesn't seem to mention this variant form directly, but seems to mention something similar as a theoretical etymology:
WH rightly notice that a nominal formation *legʰ-to- 'bed' would be strange, but it does not seem impossible: 'lied upon' > 'bed'. The form could be interpreted as a nominalized verbal adj. with regular e-grade in the root, [...] An alternative solution is to posit a tu-stem *legʰ-tu- 'the lying', which could have switched to the o-stem inflection by the time of Plautus.
While I was looking for more information about the possibility of lectus declining as a "u-stem" or "fourth declension" noun, I found an interesting discussion where some people suggest that lectus, lectus could be understood in certain contexts in post-Classical Latin as an abstract -(t)us noun derived from lego. I don't know if that might be relevant to the formation of lectuarius; I would suppose not, but I'm not completely sure.
What is going on with lectuarius—why isn't it lectarius instead? Are there any examples of similar formations with other roots where the adjective ends in -uarius even though the noun declines as an o-stem?
Edit: I found an article by Oscar Brugman that seems to discuss the declension of lectus, "Observations Plautinae et Terentianae", but it is in Latin so it's hard for me to read it. If anyone could write an answer that simply summarizes and explains what this source says, I would appreciate it. (I found this source via a citation in "Jahresbericht über die Fortschritte der klassischen Altertumswissenschaft", 1873.)
Brugman mentions one example of lectuariam and one example of lectualia:
necnon pro nobis faciunt duo verba a stirpe ‘lectu’ derivata lectuarius et lectualis, quae ita ut de scriptura dubitari nequeat leguntur apud Nonium 537,20: “plagae grande linteum tegmen quod nunc torale vel lectuariam syndonem dicimus” et in Possidii vit. Aug. cap. 22 : “vestes eius et calceamenta vel lectualia ex moderato et competenti habitu erant.” praeterea persaepe media aetate voces ‘lectuarium’ et ‘lectuale’ usurpatae sunt, quorum locorum multitudinem praebet Du Cange et Carpentarii gloss. med. et inf. Lat. ex omnibus autem quae protuli sequitur ut lectus non modo a Plauto Terentioque quarta declinatum sit, sed per totam latinitatem in usitato quidem sermonis genere, neque dubito quin qui in scriptoribus legendis hoc curent etiam alia exempla huius usus inventuri sint.
Edit: I found one other example that seems similar: annualis corresponding to annus, anni. The Lewis and Short entry says "a year old (post-class. and rare)".