After the loss of vowels at morpheme boundaries in both Sindarin and Noldorin, where a nasal appeared before a voiceless spirant (f, th, ch, s) and formed clusters of three or more consonants, then that nasal vanished. This sound change was noted by David Salo, though he more conservatively assumed the change was limited to clusters ending in a liquid r, l (GS/§4.113). Tolkien himself hinted at the change in phonetic notes from 1962 (published after Salo’s Gateway to Sindarin):

As in Sindarin general[ly] m, n tended to be lost before another consonant, camprû > camfrū > cāfru (PE17/131).

This note leads me to believe such nasal-loss before voiceless spirants was fairly common, but all the attested examples of such nasal-loss in three-consonant clusters involve a liquid, as in Salo’s phonetic rule (GS/§4.113). There are a number of examples of this phonetic development in both Sindarin and Noldorin:

• ✶ninkwiraite > ni(m)phraed > S. niphred “pallor” (PE17/55), also N. nifred of similar origins (Ety/NIK-W).
• ✶tanklā > tankla > tachl > N. tachol “clasp, brooch” in both the 1930s and around 1950 (Ety/TAK, PE18/100).
• ✶kwentrō > pentro > N. pethro-n “reciter” in both the 1930s and around 1950 (Ety/KWET; PE18/50, 100).
• ᴹ✶la(n)sro-ndo > N. lhathron “listener” (Ety/LAS²), with later cluster sr > thr.

There is one Noldorin example that shows nasal-loss before only a single voiceless spirant:

• ON. inpanta > in-pant > N. ifant “aged” (Ety/YEN), also reappearing later in the name S. Torhir Ifant (WJ/192).
• @@@ address S. iphant and ephed- < AT. enpet- with ph = [ff], compare nasal mutation and the rule where [n] assimilated to following labial at morpheme boundaries.

This nasal-loss was not the usual pattern with homorganic nasals followed by voiceless spirants, since normally medial [mɸ], [nθ], [ŋx] became [mm], [nn], [ŋg]. It is likely the dental [n] vanished in this case because it was not a homorganic nasal with labial [f]. It is tempting to assume that homorganic nasals with spirants first became long nasals, and then any remaining nasals vanished before spirants whether they were part of a cluster or not. Unfortunately, this cannot be the case either, since homorganic nasals did vanish in clusters of three or more consonants, as was the case in many of the examples above: ni(m)phraed > S. niphred, etc.

My best guess is that nasal-spirant combinations changed in three waves: (1) nasals vanished before spirants in clusters of three or more consonants, (2) remaining homorganic nasal-spirant pairs become (voiceless) long nasals and finally (3) the nasal vanished in all surviving non-homorganic nasal-spirant pairs.

Conceptual Developments: Lack of examples makes it hard to determine whether this sound change occurred in the Gnomish of the 1910s, but there are at least a couple examples of something like this phonetic development in the Early Noldorin of the 1920s, though at this conceptual stage it seems the nasal vocalized to [i]:

• ᴱ✶tank’rú > ᴱN. taichr “prop, support” (PE13/153).
• ᴱ✶kampru > ᴱN. caifr “flea” (PE13/140; PE14/66).

Coincidentally the primitive form of the last example is essentially the same as the first Sindarin example quoted above, written over 30 years later.

GS/§4.113 @@@