It is well known that non-initial y ([j]) became vocalic [i] in both Sindarin and Noldorin. This is reflected in its spelling system, where i represented a vowel in the middle of words, but y at the beginning of words:

I initially before another vowel has the consonantal sound of y in you, yore in Sindarin only: as in Ioreth, Iarwain (LotR/1114).

This shift of [j] to [i] took place in two phases (or three phases if you include Primitive Elvish developments: see below). First, any y appearing between vowels became i ([VjV] > [ViV]), and later any remaining non-initial y became i ([CjV] > [CiV]). The timing of two phases is most obvious from Old Noldorin examples. Compare:

• ᴹ√GÁYAS > ON. gaia “dread” (Ety/GÁYAS).
• ᴹ✶wāyā > wōia > ON. uia “envelope” (Ety/WAY).
• ᴹ✶beuyā́- > ON. buióbe “to serve, follow” (Ety/BEW).

Versus:

• ᴹ√MIR > ON. mirya “precious” (EtyAC/MIR).
• ᴹ√PHIN > ON. phinya “skillful” (Ety/PHIN).
• ᴹ√SIR > ON. sirya- > N. sirio “to flow” (Ety/SIR).
• (but note possible exception ON. skhalia- “to screen”, Ety/SKAL¹)

The only attested example of intervocalic y in Old Noldorin is Phayanṓr (N. Feanor), and this probably represents an older form. Also note that there would be no examples of pre-consonantal y [VjC] in Old Noldorin or Old Sindarin, since these had already become i in Primitive Elvish, and in fact were the main source of primitive i-diphthongs: [VjC] > [ViC]; this primitive sound change makes for “three phases” of y becoming i, as mentioned above.

The picture is less clear in Sindarin, but there is indirect evidence that intervocalic y became i in Old Sindarin, because of how the resulting diphthongs developed. In particular, the resulting i-diphthongs underwent the same development as original primitive i-diphthongs, indicating an early phonetic change. For example:

• √WAYA > S. gwae “wind” (PE17/33, 189).
• ✶aþayā > S. athae (“?healing”) (PE17/49).

This indicates the change of [VjV] > [ViV] probably occurred at an ancient stage, since otherwise the resulting [ai] diphthong would have survived. Diphthongal developments in Noldorin also support such an ancient change, such as: ᴹ√GÁYAS > ON. gaia > N. gae “dread” and not **gai or **gei.

Finally, there is one Old Noldorin example that indicates that the sound change [-Vj] > [-Vi] (final y after a vowel becoming i) occurred only in Old Noldorin, and not in Primitive Elvish:

• ᴹ✶nāyǝ > noi > ON. nui “lament” (Ety/NAY).

Here the final indeterminate vowel [ǝ] (probably originally short [ă]) was lost primitively, but the result was not a primitive diphthong [āi] > [ai], which would have become [ae] or [oe] in Noldorin. Instead, it seems that the long [ā] became [ǭ], and then the resulting diphthong [ǭi] ultimately became [oi] > [ui]. This is only possible if the final [j] was preserved long enough for these Noldorin-only phonetic developments to occur.

It isn’t clear whether this preservation of primitive final [j] after a vowel applied to (Old) Sindarin’s phonetic history, but for now I am assuming it does until contrary evidence appears.

GS/§4.32 WGHC/§100i-2 [usually remained in Welsh between vowels] @@@ after diphthong changes but before oi > ui