In Eldarin languages in general, where a long vowel appears before a consonant cluster, that long vowel was shortened. This was true in Primitive Elvish, but also applied to long vowels that came into contact with clusters after Quenya-only sound changes like the Quenya syncope. For example:

• Val. Mānawenūz > Mánwen > Q. Manwe (WJ/399).

It also applied to long vowels before consonant cluster from grammatical inflections such as pronominal suffixes:

• ✶ mat-ubā-ni/njē > Q. matuvanye “I will eat” (PE22/132).

Likely this remained an active sound change throughout the history of Quenya, but there are a few exceptions to this rule. In particular, a long vowel could survive before certain palatalized and labialized pairs such as ty and qu [kw], as discussed in the entry on how [j], [w] often became [i], [u] after consonant groups, for example: mátya (PE19/48, 100) or láquet- (PE22/167). Not coincidentally, both ty and qu are among the valid initial clusters in Quenya. Long vowels seem to survive before many of these clusters, such as ványe, válye “I won’t, you won’t” (PE22/162). But note the examples above of reduction before the clusters nw, ny, both of which can appear initially in Quenya.

Some scholars of Elvish languages think that these long-vowel-survivals indicate these initial and medial groups were actually unitary consonants: ty = [tʲ] or [c] rather than [tj], and qu = [kʷ] rather than [kw]. Tolkien himself generally referred to these combinations as “groups”, however (PE19/37, 80). Either way, the difference in pronunciation would be minimal. See [j], [w] often became [i], [u] after consonant groups for further discussion.