In Sindarin a dental spirant th [θ] or dh [ð] dissimilated (usually to s) away from another dental spirant, usually one that preceded it. The clearest example of this is úthaes where Tolkien specifically noted the dissimilations:

usahtie “inducement to do wrong”; S. úthaeth > dissimil. úthaes {úsaeth} (VT44/30).

Tolkien wrote (and deleted) úsaeth after úthaes, indicating he was unsure which th would dissimilate, but eventually decided it was the following th. There is another example of this dissimilation with the feminine suffix -eth:

thindā. Q thinda/sinda. S thenn. ... Masculine singular ending referring to persons or animals was -on, or after n -or: thennor, pl. -yn, -yr. Feminine -eth, after ð, th -es, pl. -ith, -is [emphasis added]. thennon, thenneth/s, thennyn. pl. †thinn, thendrim, thennath (PE17/141).

There are other examples of similar but distinct dissimilations:

• Aranrúth “King’s Ire” rather than **Aradhrúth, where the usual sound change of [nr] to [ðr] was inhibited by the final spirant (S/201).
• Hadhod “Dwarf” [< chaðǭd < khazād] rather than **Hadhodh, where the usual sound change of the final [d] to [ð] was inhibited by the preceding spirant (WJ/388).
• Edhellos “Elven-flower” (PM/346) and mallos “golden flower” (PE17/100) [< *malthǝloth] where the th in loth “flower” dissimilated away from a preceding dh or th.
• [N.] Hadhafang “Throng-cleaver” the result of dissimilation of older †Haðathang, where th became f; Tolkien also considered Havathang (Ety/STAG).
• sammath “chambers” could also be the class plural of [N.] tham “hall” with dissimilation of its initial consonant (LotR/942).

Aranrúth and Hadhod are dissimilations of a preceding dh, but probably represent an inhibited sound change rather than an induced change. The Noldorin example Hadhafang shows a change of th to f (or dh to v) rather than to s, perhaps because the sounds were intervocalic. The example mallos [< *malthǝloth] indicates the dissimilation was relatively early, before the lth became ll; hat-tip to Ellanto for pointing out this last idea to me (originally deduced by Vyacheslav Stepanov and Elaran).

Regarding the possible example sammath, it is more dubious because there are other examples where the class plural suffix -ath does not cause dissimilation, as seen in the class plural form thennath given above (PE17/141), as well as thoronath the class plural of thoron “eagle” (S/243). Thus it seems this sound change was no longer active in Sindarin, and affected only older words, not newer inflections.

Conceptual Development: Words both beginning and ending with th are fairly common in Gnomish, so it seems unlikely this sound change applied to the earliest conceptual stages of the language. Whether it affected Noldorin is unclear; I have yet to find an example one way or the other.

The úthaes example was pointed out to me by Elaran and the thennes example by Fiona Jallings, which originally prompted me to write this entry. Hat tip to Elaran, Vyacheslav Stepanov, and Ellanto for providing the other examples.