Tolkien mentioned that the nasal velar [ŋ] vanished in Sindarin with compensatory vowel lengthening on at least two occasions, both in conjunction with the etymology of S. têw or tîw “letter”:

S ñm = ¯w. Q ñ fell in with g, h, but S. ñ vanished early with lengthening. [tekma > teŋ̃ma > tēw earlier in the page] (PE17/44).

[in a chart of the phonetic development of nasal combinations] ŋn > ŋn > ¯n; ŋm > ŋw̃ > ¯w; [teŋmā > teŋwā > tēwa > S. tîw earlier in the page] (PE22/149).

Relatively few velar nasals [ŋ] survived past the Old Sindarin period. In Primitive Elvish [ŋ] either (a) assimilated to following stops and aspirates becoming [n] or [m] or (b) it often became [ɣ], such as before liquids [r], [l] (PE18/102, PE19/98). Furthermore, in Old Sindarin, any initial [ŋ] became [ŋg] or [g]. Ultimately, the only Sindarin survivals of [ŋ] were [ŋg] and [ŋk] (PE18/104), but based on the notes above it also survived for a time in combination with other nasals (except [ŋŋ] which became [ŋg]) before ultimately vanishing with compensatory lengthening as described above.

Conceptual Development: These Sindarin vocalizations of [ŋ] seems distinct from the Noldorin phonetic developments, where the vocalizations of [ŋ] before primitive nasals [m], [n] aligned with the vocalizations of [ɣ]. In part this was because in (Old) Noldorin all stops became nasals before nasals, so that primitive [ɣn], [gn] and [ŋn] all shared the same vocalizations (to [in] or [un]) in the Noldorin conceptual stage, and similarly for [ɣm], [gm] and [ŋm]. This was not the case in Sindarin, where primitive [gn] and [ŋn] had distinct develoments (to [in] or [¯n], respectively) and similarly for [gm] and [ŋm]. There are some remnants of the Noldorin phonetic developments ([eŋw] > [eiw] rather than [ēw]) in an etymology of S. têw appearing in notes from the 1950s:

• ✶teñwa > teiw > S. tew “letter” (PE17/44).

But this was soon replaced with the etymology given above where [ŋm] > [ŋw] > [¯w]. In later writings Tolkien does seem to have vacillated on whether this ŋ-vanishing occurred before or after long [ē] became [ī] in (Old) Sindarin. If before, the Sindarin word for “letter” would be tîw ([eŋw] > [ēw] > [īw]); if after the word would be têw. The plural form “letters” was tîw in both cases. In The Lord of the Rings proper, only the plural form t(h)iw appeared: Celebrimbor o Eregion teithant i thiw hin “Celebrimbor of Hollin drew these signs” (LotR/305). Thus, Tolkien’s vacillation is understandable; either paradigm was consistent with the published material.

Neo-Sindarin: For Neo-Sindarin purposes, I think it is best to assume [ŋw] > [¯w] after [ē] became [ī], since têw “letter” is better attested and is probably more widely accepted than singular tîw.