Just as long vowels shortened before clusters, it seems that in both Sindarin and Noldorin the long vowels [ī], [ū] often shortened in polysyllables, as noted by David Salo (§4.166). Tolkien mentioned this phonetic development in his notes on the Noldorin usage of the Feanorian Alphabet from the 1930s:

... thus the ending au, ī, ū becoming thus unstressed was later reduced to o, i, u ... (PE22/31).

Lack of stress seems to be the main trigger for this sound change, and it mainly applied to long vowels in the final syllables of polysyllables (which were rarely stressed). The clearest examples are in Noldorin:

• ᴹ✶at-yēn-ar [> *edīnar] > N. edinar “anniversary day” (Ety/YEN).
• ᴹ✶Ithīl > N. Ithil “Moon” (SD/306).
• ᴹ✶kwelēnē [> *pelīn] > (soft-mutated) N. belin “*fading” (Ety/LAS¹).
• ON. wanūro [> *gwanūr] > N. gwanur “kinsman” (Ety/NŌ, THEL, TOR).
• ᴹ√RŌ [> *rhūmen] > N. †rhufen “east” vs. ᴹQ. rómen (Ety/RŌ).

As the last two examples illustrate, since short [u] often became [o] in Sindarin and Noldorin, many examples of short [u] were originally long [ū] that had shortened. This is one of the ways [u] could survive in Noldorin, since in Old Noldorin long [ō] became [ū], and similarly for Old Sindarin. This is demonstrated in the clearest Sindarin example of the sound change [ū] > [u]: S. onur past tense of S. nor- “to run” (PE17/168): the u must have been from long [ū] because otherwise it would have become o. A similar shortening is seen in S. eniðen “*I meant to do it”, 1st sg. past tense of S. nidh- “to will, intend” (PE22/165). But this shortening is not seen in all past tenses: S. edíw past tense of S. dew- “to fail” (PE17/151).

The shortening sometimes happened even when [ī], [ū] were stressed, but these seem to be exceptions:

• ᴹ√MŌ [> *mōda] > N. muda- “labour, toil” vs. ᴹQ. móta- (Ety/MŌ).
  • Compare: ᴹ√LUK [> *lukta > *lūtha] > N. lhûtha “to enchant” (Ety/LUK; EtyAC/LUK).
• ᴹ√RŌ [> *rhūmen] > N. †rhufen “east” vs. ᴹQ. rómen (Ety/RŌ), but later S. rhúnen.
• ᴹ✶neiniel- [> *nīniel] > N. niniel “tearful” (Ety/NEI), but later S. Níniel.

There are numerous examples of long vowels in stressed positions that were not shortened:

• S. dínen “silent” (PE17/98, S/121).
• S. fíreb “mortal” (WJ/387).
• S. and N. †ínias “annals” (MR/200, Ety/YEN), but in one place N. inias (LR/202).
• S. írui “desirable, lovely” (PE17/155).
• S. míria- “to sparkle like jewels” (PE17/24).
• S. Deldúwath “Deadly Nightshade” (S/155, WJ/282).
• S. Drúadan “Wose” (PM/324, UT/385).
• S. rhúnen “eastern” (S/124; SA/rómen).
• S. Túrin (S/198).

There are a few long vowels that were preserved even in unstressed positions:

• S. Curunír “Man of Skill” (LotR/1085).
• S. Elurín, Eluchíl (S/188, 234).
• S. Glanhír “Boundary Stream” (UT/318).
• S. Nauglamír “Necklace of the Dwarves” (UT/318).
• S. Aranrúth “King’s Ire” (S/201).

My best guess is that where it was part of a recognized element of a compound, the long vowel might have remain or been restored by analogy with the independent form. It’s unclear whether the vowel shortening applied to unstressed [ȳ], since we have no clear examples of this long vowel in an unstressed position. It was probably not mentioned in Tolkien’s original note above because long [ȳ] did not naturally arise in Noldorin, except as the lengthening of [y] in monosyllables (PE22/38). Given the comparative rarity of clear Sindarin examples, however, it’s possible that this vowel shortening was not a regular part of Sindarin phonetic development at all, even for [ī] and [ū], being only regular in Noldorin.

For a discussion of the possible shortening of [ǭ], see then entry on how [ǭ] became [au].

§4.166 @@@