Q. spirants became stops after nasals; [mɸ|nθ|ŋx|nð] > [mp|nt|ŋk|nd]
In Quenya anytime a nasal came into contact with a voiceless spirant in the Parmaquesta [PQ] period, the spirant was restopped. Tolkien mentioned this development in both the Outline of Phonetic Development [OP1] from the 1930s and the Outline of Phonology [OP2] from the 1950s:
But other cases occur (a) where nasal was reintroduced after opening of spirant: mꝑ, nþ, ñχ then result in mp, nt, ñk; and (b) reintroduction producing TQ nth, ns (OP1: PE19/44).
In some later formations where the nasal was reintroduced, after the aspirates had become fully spirantal in unnasalized forms but had not yet shifted point of articulation the nasal was not unvoiced, but the spirant was closed to a stop, producing mp, nt, ñk. Where this grammatical renewal occurred in TQ after the change of þ > s, ns remained as in TQ panse “smoothed” ... No cases of renewal before f, h occur (OP2: PE19/89).
This change would not occur for nasals + aspirates originating in Primitive Elvish, since nasals became voiceless stops before aspirates in Ancient Quenya. As the notes above indicate, the change also did not occur in Tarquesta [TQ], where spirantal f, h were preserved, or in the case of þ developed into s as usual. Such combinations are occasionally attested in late [TQ] compounds such as henfanwa “eye-screen” (PE17/176).
Thus this restopping only applied to combinations formed in PQ, in the window when these spirants were still ꝑ, þ, χ [ɸ, θ, x] and had not yet undergone their changes to TQ f, s, h. The only possible examples of this sounnd change that I’ve been able to find are in some tentative past tense forms from the late 1940s in (rejected) rough notes associated with Tolkien’s discussion the Quenya Verbal System:
EPH-. (especially from water, opp[osite] of “dive”) “emerge”. ephá “it emerges”. empe. emphe. [In the left margin:] ephe, emphe. ethe, enthe. ehe, ēhe. eñghe; enche (PE22/127).
Of these, emphe [>] empe seems to represent restopping. However, elsewhere Tolkien said such combinations in past tense forms normally retained their Ancient Quenya development to pp, tt, kk: rappe past tense of raf- from √RAPH; lakke (or lahte < *lakhne) past tense of lah- from ᴹ√LAKH; pitte past tense of pis- from ᴹ√PITH (PE19/44, 89; PE22/102-3), though in the case of þ > s the past was often reformed to ns, as in the TQ past tense form panse of pas- from √PATH as given above (PE19/89). There is another apparent example of restopping in The Etymologies from the 1930s:
However, the related word ᴹQ. makar “tradesman” (rather than **mahar) suggests that this is more likely to represent a variant form of the root: *√MBAK.
There is one example which hints at a similar restopping for voiced spirants after nasals, but again such combinations would be very rare:
Conceptual Development: In the Early Qenya of the 1910s and 1920s, stopping of voiceless spirants after nasals was the normal development in all cases:
all nasals + x͡w > nq ... nasal + χ > nc ... nasals + hy [IPA ç] > nty ... nasals + þ > nt ... nasal + ꝑ gave mp (PE12/17-20).
The same is true of voiced spirants, with the sole exception of ʒ̑ [IPA ʝ] > y:
ñ, n, m, ŋ + ẇ [>] ngw ... n, m, ñ + γ > ng ... ʒ̑ [>] y. n, m, ŋ, ñ + y [>] ny ... rđ, nđ, lđ, zđ > rd, nd, ld, rd ... rƀ, lƀ, mƀ, zƀ > rb, lb, mb, rb (PE12/23-24).
There are numerous examples of these sound changes in the Qenya Lexicon from the 1910s. Tolkien's conception of the voiceless phonetic developments seems to have shifted after he introduced aspirates into Primitive Elvish in the 1930s.