The endings -on and -ion are quite frequent in names, but their role is not always clear. So I will try to treat them more carefully.
Already in the earliest Celtic-style language Goldogrin both -on and -ion are adjectival endings (beside -n, -in), e.g. argulthion ’equal, equivalent’ (GL:20), gwedhwion ’looped, bending’ (GL:46), martion, mart ’fated, doomed, fey’ (GL:56), taithion, godaithion ’educated’ (GL:68) or falon, falin ’naked’ (GL:33), helon ’frozen’ (GL:48), hebon ’bound – also bounded, surrounded’ (ibid.), malon ’yellow’ (GL:56), melon, meltha ’dear, beloved’ (GL.57).
At the same time -ion is the genitive plural of consonantal nouns, e.g. glôr ’gold’ > glorion; and -on is the genitive singular of nouns ending in -a/-u, as coma ’disease’ > comon, culu ’gold’ > culon (GL:12-14).
By the time of The Etymologies Noldorin, Ilkorin and Doriathrin are the languages of the Celtic branch. Both -on and -ion occur as agentive suffixes, as in #faron ’hunter’ (SPAR-), Dúrion/durion ’a Dark-elf’ (DOƷ-, DÔ-, MOR-) or else form names as Mirion ’ordinary N name of the Silevril (Silmarilli)’ (MIR-), Gelion ’merry singer’, also a river (GYEL-). Probably associated with this is the patronymic suffix -ion < YŌ, YON- ’son’. Tilion ’hyrned’ or ’the Horned’ (TIL-), Brithon ’pebbly’ (BIRÍT-) and Erchamion ’one-handed’ (beside Erchamui, Ermabuin, Ilk. Ermab(r)in (MAP-, LR:427,146,405)) look like adjectival forms but are also names at the same time. N. Erchamron and later S. Erchamon ’one-hand Man’ (VT47:7) are not adjectives. Purely adjectival seem to be Ilk. gelion ’bright’ from GAL-, tovon ’lowlying, deep, low’ < tubnā (TUB-).
At the same time Ilkorin shows the genitival inflection sg. -a, pl. -ion as in Dor-thonion ’Land of Pines’, Torthurnion ’King of Eagles’ [4]. The genitive pl. -ion is probably also found in Noldorin in Eredwethion ’Mountains of Shadow’ (TI:345, WATH-). This is explicitly Noldorin, the Ilkorin variant being Urthin Gwethion. However, another translation is ’Shadowy Mountains’ (LR:447) hinting at an adjective #gwethion ’shadowy’. Also Duil Rewinion ’Hills of the Hunters’ (LR:286).
But that is not all yet – the simple addition of -ion seems to denote a region: Dor. Regornion ’Hollin’ (ÉREK-) from regorn ’holly’.
The suffix -on (and -ion for i-stems) is also often augmentative: N. annon ’great gate’ (AD-), later S. (g)aearon ’the Great Ocean’ (Rgeo:72-73, PE17:27) < aear ’sea’, Sirion ’the Great River’ < sîr ’river’ (Silm.index).
See also [5].
The close association between patronymic -ion and gen. pl. -ion that also occurs in Quenya and is given account in the Early Qenya Grammar:
-ion, old patronymic ending, which has appearance of being a genitive plural and hence is often formed from -li form of vocalic nouns, as Noldolion (pl. noldoliondi) ’descendant of the Gnomes’ (PE14:45).
In Noldorin/Ilkorin names -ion is indeed of a manifold ambiguity. So how should it properly be: Dufinnion *’dark-hair-person’ or adj. *’dark-haired’; Annerchion *’gate of goblins’ or *’goblinish gate’; Torfirion *’high-man + name-suffix’ or maybe *’high-men-place’, Andon *’long one’ or adj. *’very long’, Amarthon adj. *’doomed’, *’greatly doomed’ or *’doom + name-suffix’, Duil Rewinion *’hills of hunters’, *’hunting hills’, *’hills of the hunting region’, Eredwethion ’Mountains of Shadow’ (gwath, gen. pl. #gwethion), ’Shadowy mountains’ (adj. #gwethion) or even *’mountains of the shadow-region’?
It will not be surprising that Tolkien often adjusted the conceptions and interpretations when it came to explanations.
Thus according to one explanation Eregion ’Hollin’ (cf. ’holly-region’ in RC:772) < S. ereg ’holly’ and Nanduhirion are said to add the regional ending -ion (PE17:42) which is the adjective iaun ’large, extensive, wide’. Related are older -ian(d) in Beleriand and pl. -iend, ien often used of a single varied region as Anórien, Ithilien. In older names it usually applied only to a large feature, as in Sirion ’the Great Stream’ (PE17:42).
Another explanation gives -ion < -ı̯aun, from yānā < √YANA ’wide, large, extensive’; also S. iaun ’roomy, wide, extensive’ (ibid.). It was applied especially to topog[raphical] features of large extent, especially long, wide rivers, long (and wide) ranges. So Sirion < siriānā; Eregion, Erydweithion, -ian. Compare essentially the same √YAN- ’vast, huge’, untranslated S. iaun as part of Rhovanion ’Wilderland’ (PE17:99).
But then Tolkien decides that Eryd-weithian(d) should = mountains of the region of the shadows. But then gweithian ’region of the shadows’ should remain unlenited in genitival position. Hence Tolkien concludes: Better return to Eryd-wethrin, shadowy mountains (i.e. with an adjectival suffix). This is how the mountains appear in The Silmarillion – Ered Wethrin
He further decides: Delete entirely yondo = ’son’! Very unsuitable (PE17:43) and comes up with a new etymology: √YŎNO ’wide, extensive’, in regional names yonde, ionde > -ion, yon. Often associated with genitive plural [Also confused with -on, augmentative or male suffix]. So now Sirion is properly ’the Vale or lands about the River Síre’ or ’the great stream’.
A yet different root is √YOD- ’fence, enclose’ yields yonde ’any fairly extensive region with well-marked natural bounds (as mountains or rivers)’ > -yonde, yon / ionde, ion frequent in regional names.
The genitive plural is just briefly mentioned here, but is explored in other notes. Thus Sindarin has gen. sg. -a, pl. -on called purely possessive (PE17:97). For the plural both endings may be combined with the ordinary pluralization via -i: lais galaðon or lais geledhion ’the leaves of trees’, similarly glim maewion or glim maewia ’(the) voices of gulls’. See also elenath ’the (host of all the) stars’ > full genitive elenathon (PE17:24-25).
A remark reads: ion is ia < g[enitive] iōm, later n [?restored]. So presumably the plural marker -n is (for some reason – perhaps by Quenya influence?) restored after having been regularly lost -iōm > *-iōn > *-io > (unstressed?) -ia. (cf. VT42:14, VT47:24 and ai-lin- > N. oel ’pool, lake’, pl. oelin < *ai-lini (AY-)).
At the end, however, Tolkien rethinks it: X DON’T have inflected genitive!
Other Sindarin samples with -on, -ion include:
Different explanations can be found here as well. According to one Galadon is actually a lenited adjective Caladon < calatāna (PE17:84). According to another Galad and especially the apparent genitive plural Galadon ’of trees’ are not Sindarin (PE17:51), but rather Nandorin with galadā > galad (S. galadh). Although the spelling was changed to Galadhon with proper Sindarin dh, caras is assigned to Silvan speech in UT:257. So the whole name should be perhaps regarded at least as dialectal, if not pure Nandorin. The Gladden Fields also lie to the east of the Misty Mountains and fit into this scheme.
From Let:347 we learn that Orbelain was a ’phonological’ translation by the Noldor. So all week days, including Orgilion could be Quenya-influenced. Orbelain includes the reconstructed adjective belain *’of the Valar’ (Q. Valanya), so maybe gilion is an adjective as well, influenced by the gen. pl. -ion in Quenya. Oraearon might also contain an -on adjective, corresponding to Q. Eärenya, or else S. (g)aearon ’ocean’ (PE17:27).
This is also supported by the gloss of Dorthonion as S. Noldorized.
In Nan Gondresgion the position of Gondresgion as a qualifier as well as the Quenya version employing the possessive case suggest that the purely possessive genitive has been intended here.
On the other hand, the sole compound Dorwinion and its translation ’Young-land country’ suggests an analysis dor-win- ’young-land’ (lit. ’land-young’) + -ion ’country’. In any case a genitive plural is not appropriate here. The river name Gwinion follows the pattern of Gelion, Tilion, Mirion and might itself be an adjective *’young’.