In both Sindarin and Noldorin, the long consonants [ll], [ss] and [nn] generally shortened at the end of polysyllabic words. Helge Fauskanger examined these phonetic developments in detail in a pair of articles, To SS or not to SS and The Question of nd or n(n) (on the Ardalambion website). Tolkien himself mentioned these developments in The Lord of the Rings appendices:

Note that consonants written twice, as tt, ll, ss, nn, represent long, ‘double’ consonants. At the end of words of more than one syllable these were usually shortened: as in Rohan from Rochann (archaic Rochand) (LotR/1115).

The combination tt was not relevant to Sindarin, since it spirantalized to [θθ] (although later the long [θθ] did shorten to [θ]), but the other three final combinations did reduce in both Sindarin and Noldorin. There are numerous examples:

• ✶galata-rig-elle > S. Galadriel “(Lady with) garland of sunlight” (MR/182, 470).
• ᴹ✶Tindōmiselde > ON. Tindūmhiell > Tinnúviel > N. Tinúviel “Daughter of Twilight” (Ety/TIN), and similarly in Sindarin.
• S. †Rochand > Rochann > S. Rohan “Horse-country” (LotR/1115).
• ᴹ✶dōmilindē > dúlind > dúlin(n) > N. dúlin “nightingale” (SD/302; Ety/DOƷ, TIN).
• S. †Caran-rass > S. Caradhras “Redhorn” (LotR/1113).
• ᴹ✶tyulussē > N. tulus “poplar-tree” (Ety/TYUL).

Reduction of -ll: Of these changes, it seems that -ll > -l was comparatively early, as seen above in the development of Tinúviel; the relative earliness of this change was noted by David Salo (GS/§4.205 vs. GS/§4.213). This sound change was also applied universally: there are no examples of final -ll in Sindarin except for monosyllables and S. Bar-goll, which doesn’t seem to be a full compound.

Reduction of -ss: The change of -ss > -s was also likely somewhat early. There are no Sindarin cases of final -ss in polysyllables except for the pseudo-compound S. Taras-ness. However, as noted by Helge Fauskanger in his article To SS or not to SS, there seem to be some examples of -ss > -s even in monosyllables:

• S. tas “index finger” vs. Q. tassa (VT48/5).
• ✶gassā > N. gas “hole” vs. ᴹQ. assa (Ety/GAS).
• N. glas “joy” (Ety/GALÁS) vs. N. glass (Ety/BOR) and ᴹQ. alasse (Ety/GALÁS).
• S. nos(s) (PE17/169) and S. nos (PM/320) “kindred, family” vs. Q. nossë (PM/320, Ety/NŌ), but note N. noss (Ety/NŌ).
• S. los “snow” (PE17/26, 161) vs. S. loss (Let/278; RGEO/62; VT42/18) and Q. lossë (PE17/26, 161; RGEO/61).
• N. rhas “horn” vs. ᴹQ. rasse (Ety/RAS), but note S. ras(s) vs. Q. rassë (PE17/36).
• S. las(s) (Let/282, RC/760) and S. las (PE17/62; PE22/166) “leaf” vs. Q. lassë (Let/282; PE17/62), but note N. lhass (Ety/LAS¹) and S. lass (PE17/49).
• ᴹ✶risse- > N. rhis “ravine” beside N. rhess (Ety/RIS²).

Fauskanger’s conclusion (with which I agree) is that these were likely orthographic variations, and the normal written form in monosyllables should be -ss, which is much more common (there are dozens of monosyllables that have -ss forms but no -s form). Of interest, however, is S. lais plural of S. las “leaf” (PE17/62, 97), and S. rais plural of S. ras “horn” (SA/ras). Typically i-intrusion did not penetrate consonant clusters, so the plurals of lass and rass would be normally be **less and **ress; compare S. nern plural of S. narn “tale” (MR/373). This hints that the despite the common spelling, the actual pronunciation was [s] rather than [ss] even in monosyllables, and this reduction took place earlier than i-intrusion.

A possible counter-example to this (hypothetical) change in pronunciation is Yssı̯ < ✶Ossai, an archaic Sindarin name of Ossë that survived in the compounds Yssion and Gaerys (WJ/400). As discussed elsewhere, in stressed monosyllables ending in a single consonant, [y] might become [ui] after i-intrusion, as seen in the plurals of monosyllables with over-long ô: N. duil plural of N. dôl (SM/225), S. thuin plural of S. thôn (PE17/188), S. thuil plural of S. thôl (PE17/188); here the phonetic development seems to be [-yCı̯] > [-uiC]. If the ss in †Yssı̯ was pronounced [s] and it followed this same development after i-intrusion, we might expect to see **Uis.

However, Yssı̯ was an older form of Ossë’s Sindarin name, and Tolkien said that it survived only in compounds (WJ/400). The stress patterns that produced the diphthong [ui] in monosyllables with over-long vowels would not have occurred in the polysyllabic compounds Yssion and Gaerys. Furthermore, if the sound shift of [ss] to [s] in Sindarin monosyllables was comparatively late, any change of [s] to [h] and its eventual disappearance could not have occurred, since intervocalic [s] became [h] back in Old Sindarin, well before the reduction of [ss] to [s]. Thus the hypothetical reduction of [ss] to [s] before i-intrusion in both polysyllables and monosyllables nicely explains the plurals S. lais and rais of S. las(s) and ras(s), even if this sound change was not usually reflected in spelling.

Reduction of -nn: There is a comment from Tolkien in his notes on the Feanorian Alphabet from the 1930s which states that the reduction of [-nn] > [-n] was in fact universal for pronunciation, but nn was retained in the spelling of monosyllables. This is further (indirect) support for the idea that -ss might be pronounced [-s] in monosyllables as well:

This [nn] became [n] finally and before consonants, and this is recognized in spelling, except that 5 = nn is retained in stressed monosyllables as a sign of the short quality of the preceding vowel. Thus sh5 = gŏnn, Gondolic gond “rock” (PE22/35).

Unlike ll and ss, there are a number of Sindarin and Noldorin polysyllables with final -nn without a corresponding -n form:

• S. aerlinn “*holy song” (RGEO/62; VT44/24).
• N. ambenn “uphill” (Ety/PEN).
• S. cidinn “small” (PE17/157).
• S. Lothlann “Wide and Empty” (S/123), N. Lhothlann (Ety/LAD).
• N. tuilinn “swallow, (lit.) spring-singer” (Ety/LIN², TUY).

Helge Fauskanger went so far as to suggest the adoption of this convention for two-syllable words in his article The Question of nd or n(n). However, the development for -nn > -n in polysyllables seems to be somewhat later than the other reductions of double consonants, and these could simply be archaic forms. There are likewise many polysyllables that retain final -nd, which is an even more archaic pronunciation. Since there are dozens of two-syllable words that show reduction of -n, I think it’s better to follow the same spelling conventions as -ss vs. -s: retained in monosyllables but reduced in polysyllables, which is also how Tolkien generally described them as well.

There are a few monosyllables that seem to show -nn > -n, such as glan(n) “boundary” (VT42/8), but as with -ss these seem to be rare outliers; the normal spelling seems to be -nn in monosyllables. However, the question of whether [-nn] was pronounced [-n] in monosyllables is complicated by the question of whether or not [-nd] was retained in Sindarin monosyllables: see the entry on how [mb], [nd] became [mm], [nn] for further discussion. It is pretty clear that [-nn] > [-n] in the pronunciation of polysyllables, however.

It is tempting to assume that, like [-ss] > [-s], the reduction of [-nn] > [-n] was before i-intrusion so that the plurals of such words would show (for example) a becoming ai. There is, in fact, an example that seems to support this: Periain plural of Perian “Halfling” (LotR/953; Let/427) versus its class plural Periannath (LotR/1116; Let/427). Unfortunately, there is a problem with this scenario. We know that (1) the reduction of [-nd] > [-nn] > [-n] occurred after [-mb] > [-m] (PE22/35) and (2) there seems to be an attested plural S. cem for S. cam “hand” (VT50/22) rather than **caim. This plural indicates that mb, mm was a cluster at the time of i-intrusion and therefore resisted the intruding i. It seems likely the same is true of nd, nn.

If the reduction of [-nn] > [-n] occurred after i-intrusion, how do we explain the plural Periain? There are plenty of plural forms that had only a single [n] primitively, as for example S. erain plural of S. aran “king” (PE17/40, 111) vs. Q. aran. Plurals like Periain could have appeared by analogy from such single-n forms. Note that such analogies would not have been possible for final -m(b) and -s(s), since any single-m forms would have developed a v and any single-s forms would have developed an h which ultimately would have vanished. Thus, the “plural by analogy” scenario does not invalidate the chain of reasoning discussed above whereby the plural cem indicates a cluster [mb] or [mm] at the time of i-intrusion but the plurals lais, rais indicate a single [s].

There are also some indications that nn could (rarely) reduce to n medially, as indicated by the following quote from the Feanorian Alphabet document:

In post-Exilic [nn] > [n] before a main stress and so also in initial mutation (of d); but this was not recognized usually in spelling: the mutation of 2 [d] is always written 5 [nn or nd archaically]. Thus 2]F7 dór “land”, pl. `=5®R7 i·nnýr [inȳr] “the lands” (PE22/36).

The best (and perhaps only?) example of this is the name Tinúviel from older form Tinnúviel (PE19/73, Ety/TIN). The note above indicates the medial reduction of nn > n before the main stress can also be seen in certain grammatical mutations, but there the evidence is more mixed. See soft mutation and nasal mutation for further discussion.

Conceptual Developments: Final -ll does not occur anywhere in Gnomish of the 1910s, not even in monosyllables, so it seems this reduction was universal at this early conceptual stage, frequently occurring medially as well and often lengthening the preceding vowel in the process of reduction, as noted by Roman Rausch in his Historical Phonology of Goldogrin (HGP/§2.3):

• ᴱ✶Tefíldŭ [> Tifill] > G. Tifil Prince of Cats vs. ᴱQ. Tevildo (GL/70).
• ᴱ✶nalda > nalla > G. nâla “yellow-lily” (GL/59).
• ᴱ✶oldō [> *olla] > G. ôla “cliff, seaward precipice” (GL/62).

Neither -ss or -nn occur finally in Gnomish polysyllables, but are common in monosyllables and medially, indicating a Sindarin-style reduction finally in polysyllables only. The same is true of ss in the Early Noldorin of the 1920s, but there are quite a few examples of final -nn in polysyllables, and it seems retention rather than reduction of -nn was the norm at this conceptual stage:

• ᴱN. aglann “ray of light” (PE13/136, 158).
• ᴱN. amvenn “uphill” and ᴱN. amvinn “slope” (PE13/159).
• ᴱN. dadvenn “downhill” and ᴱN. dadvinn “downhill slope” (PE13/161).
• ᴱN. eglann “very wide” (PE13/142).
• ᴱN. orvlann “without deceit” (PE13/156).
• ᴱN. urhonn “heartless” (PE13/156).

But:

• ᴱN. elvain(n) “wondrous” (PE13/161).
• ᴱN. Angban(n) “Hells of Iron” (PE15/62), in the texts usually Angband.

For reductions of ll in Early Noldorin, there are not enough examples in the 1920s to deduce a clear pattern.

GS/§4.205 GS/§4.213 GS/§4.129 @@@ shift of [ll] may have been earlier, as indicated by gannel

HPG/§2.3