The +#Story_of_a_Root G/3AR is typical and specific simultaneously. Its most difficult part having already happened before the Etymologies and most of the later complications arising from the clash of phonology altered and lexicon published. The discussion will not be long, but it will demonstrate some typical features of Tolkien's manner of work.

* Late 1930s In The Etymologies, after at least two (!) levels of alteration and rewrites, Tolkien achieves the following status quo (Etym:3AR, see also GAR, EtyAC:same): there is not one, but two presumably closely intertwined roots, 3AR (glossed 'have, hold') and GAR(-AT, -AD) (never glossed in the final version, but with the derivatives given making the difference in meaning tenuous if existent).

3AR produced in Quenya, harya 'possess' [Note 1], harma 'treasure' (a single thing), harwe 'treasury' (a set), haryon (lit. 'possession-son') 'heir > prince' and haran, harni 'king, chieftain (of a land, never a people); in Noldorin, ardh 'realm', aran 'king' (=Q haran).
GAR produced in Quenya, arda 'realm' (apparently derived the same way as N ardh but from a parallel root), aryon 'heir', arwa 'in control of, possessing' (with Gen.), suffix -arwa '-having'; in Noldorin, a verb garo, gerin 'have, hold' [Note 2] and its substantivised fossilised participial garn 'property'. The extension GAR-AT provides a base for Q arta, N garth 'fort, fortress' (the meaning shift is 'to guard, protect, keep').
Doriathrin garth 'realm' and garon 'lord' might be from both, for in this language initial [g] and [3] coincided in [g] (per PE18:103, Tolkien later shifted the result of [3] to [h], as in Quenya, but never made the necessary alterations in the vocabulary).

It is mysterious why Tolkien ever decided to introduce such a complicated interplay of two roots (most of the words here are either new or harvested from many sources of pre-Etym times, so retention of older favourites is out of the question). Perhaps, some proper names (Garthurian? Arthurien? Arthorien?) he wished to support?

However, the thing that broke the system was the alteration in the phonological basis. The developments of the Comparative Tables strongly diverge G from 3, with the former giving zero in Q, G in N, while the latter devoicing to H in Quenya [Note 3] and disappearing in Noldorin. Yet during the early 1950s changes to Outline of Phonology, Tolkien eliminated the element 3 altogether! He introduced [h] instead, with only minimal difference to the 'formerly' [3], including remaining in Quenya and vanishing in Sindarin. (Ab initio he wrote it to vanish in Telerin as well, which makes this suspiciously close to the 'Etym-[3]' — see Elwe, Etym:3EL — but in QE already H remains in both tongues of Aman, and in late 1960s Tolkien fights his desire for H to remain in Amanya Telerin only! See PE22:163 et al.) When [3] came back to CE in the 1960s, it behaves word-initially totally alike [g] in Q and S (see Notes on Óre).

* Late 1950s All if this led to no change concerning GAR with most stably developing G: and indeed, Quendi and Eldar provide Q arda and S gardh derived from gardā (QE:402). The problem comes with all the derivatives of 3AR, suddenly obsoleted.

It is supported by Tolkien's arising desire to have aran, not haran, for 'king' in Quenya as well. In two 1957 documents (QN and NN, PE17: 147, 118) he makes an alternate etymology, through the root RĀ/ARA 'excellent (not implying the rest are bad)' and its extensions AR-AT(N) and totally fixes it, giving in QE:369 Elwe, Aran Lestanórëo. (Curiously enough, he additionally uses Aran Sindaron there, despite his previously expressed insistence of such use's incorrectness.)

So, aran is taken out of reach, but we cannot merely forget the derivatives of 3AR (which still make sense as derived with *HAR instead!) I'm not talking about harya in the doubtfully dated Merin sentence, but of tengwa 11, harma 'treasure'. Knowing that the initial H can, in post-LotR Quenya, be explained only as former H, KH and SK, this HAR seems a legit proposition, allowing to retain all the former derivatives.

Significatio.

Apparently, those who wish to update their Etymologies for Quenya and Sindarin, should replace 3AR to HAR throughout and strike their (h)aran away. Otherwise, the stem(s) and their developments are still totally actual on any stage of the legendarium.

Note 1. Due to the verbality of the root meaning, it can be postulated that this in a sirya-like example of a half-strong PE22:114 verb, with tense-forms along the line of háre, harne/háre, ahárie, yet haryuva.
Note 2. The description of the original GAR entry expansion, VT45:14, provides some additional knowledge about the verb: despite forming the infinitive garo and past garant (presumably pres. and pa. parts. garol and garannen as well), it still shows basic gerin, gâr (?) in present. This could be a general pattern of Noldorin verb regularization: unlike Quenya, this language made a wide use out of an apparently meaningless a extension, possibly borrowed from the 'genuine' a verbs. Cf. MEL melo in Etym, with the analogical type of conjugation, and pp. 74-5 of Thorsten Renk's Pedin Edhellen for a more concise approach.
Note 3. Strictly speaking, both the Tables themselves and the Outline, PE19:019, 073 give zero for the Quenya derivative of [3] as well. Yet here the factual evidence of the words (cf. Etym:3ARAM, 3EL, 3Ō) themselves contradicts the theoretical postulations.

Next time: probably, something on the lines of KHAD/M/W (and sitting).