Q. [θ] became [s]; [θ] > [s]
In the Ñoldorin dialect, the voiceless dental spirant þ [θ] sound that had developed from the primitive aspirate th [tʰ] further developed into s. This sound change did not, however, occur in the Vanyarin dialect, and represents one of the most distinctive differences between the two main branches of Quenya. In fact, the change of þ > s may be the first divergence of the two dialects, since it seems to precede other divergences such as Ñoldorin [ð] > [r] (to [z] in Vanyarin) and [z] > [r] (with [z] surviving in Vanyarin).
This sound change had some interesting political ramifications as well, in particular because it involved a change in the pronunciation of Serindë, one of the names of Fëanor’s deceased mother. This may have contributed to Fëanor’s alienation from other Elves, as discussed in the The Shibboleth of Fëanor, written by Tolkien in 1968 (PM/331-339). This makes þ > s one of the best documented sound changes in all the Elvish languages, and Tolkien mentioned it frequently (Let/425; LotR/1114; PE17/129; PE19/31, 34, 49, 71; PE22/13, 44-45).
This sound change was also established very early in Tolkien conception of Qenya. He first mentioned it in the Qenyaqesta of the 1910s:
þ ... It gave s universally, except where it had already undergone certain changes from þ > t or đ or d (PE12/19).
In the Outline of Phonetic Development [OP1] from the 1930s, Tolkien attributed this sound change to the Lindarin [pre-Vanyarin] dialect:
In L. th (þ) > s. This became very general in TQ, with result that þ, s were confused in TQ both in spelling and pronunciation except among the learned. In learned and formal writing the distinction was maintained in writing; and the Noldor, who beside being the chief lore masters, also possessed the sound [þ] in their own tongue, usually represented AQ, PQ th by (þ) (OP1: PE19/31).
Note that in later revisions of OP1 Tolkien changed “L.” >> “V.” in the first sentence reflecting the name change of this dialect from Lindarin to Vanyarin (OP1: PE19/31 note #13).
In the 1930s, the Noldor retained the pronunciation þ because it was a common sound in their native Noldorin [pre-Sindarin] language. After Tolkien changed Noldorin to the native language of the Sindar, this scenario shifted and in the Outline of Phonology [OP2] from the 1950s Tolkien attributed the sound change þ > s to the Ñoldorin dialect of Quenya:
The Vanyar preserved the sound [þ]; but this became s in the speech of the Noldor not long before the Exile. This s for th > þ was thus originally characteristic of the Tarquesta, in which CE th, s coalesced in s initially. Not medially, for the spirants derived from aspirates did not become voiced, and s < þ remained voiceless medially while older s became z (see below). The PQ spelling with distinct signs for þ and s was, however, maintained and later many among the Exiles restored the sound [þ], after their adoption of Sindarin as their diurnal speech, a language which favoured the sound [þ]. Some retained it in imitation of the Vanyar: cf. under [s] and [ñ]. This was done chiefly in reciting or reading aloud matter written in PQ. In the actual use of the TQ as a spoken language s for þ remained characteristic of the Noldor (OP1: PE19/71).
Tolkien also mentioned this sound change in The Lord of the Rings Appendix E:
TH represents the voiceless th of English in “thin cloth”. This had become s in spoken Quenya, though still written with a different letter (LotR/1114).
Indeed, the name of the tengwar in question, súlë “spirit” (3) was original pronounced thúlë (LotR/1117), and was used exclusively for s of this origin, as opposed to silmë “starlight” (8 or i) used for s for other origins (including surviving primitive s).
There are some interesting questions surrounding the timing of the sound change þ > s, but I believe it occurred fairly early in Tarquesta [TQ]. In particular, in Ñoldorin TQ only, z dissimilated to s when it preceded or followed r. This is more plausible if s had already been reintroduced as a medial sound by þ > s, meaning this shift probably occurred before z and ð merged into r in Ñoldorin Quenya. In the Eldamo data model, I am using the þ > s sound change to mark the start of TQ and the beginning of the divergence of the Vanyarin and Ñoldorin dialects.
Indeed, in the aforementioned Shibboleth of Fëanor, Tolkien stated that this sound change took place before or shortly after the birth of Fëanor himself (PM/332). This scenario presents other problems though, because it means the Feanorian Alphabet could not have been invented until the TQ period. There is ample evidence elsewhere that tengwar were used in very early in the Parmaquesta period, such as the history the tengwar anna (h) originally used for lost ʒ, a loss that began in Ancient Quenya [AQ] and finished in early Parmaquesta [PQ]. The Eldamo data model assumes the introduction of tengwar marks the boundary between AQ and PQ.