In Ancient Quenya [AQ], the combination ns became ss. Since primitive nasals assimilated to n before s, this was the ultimate result of all primitive nasals + s. This sound change was mentioned in both the Outline of Phonetic Development from the 1940s (OP1) and the Outline of Phonology from the 1950s (OP2):

The nasals were frequently assimilated to dentals ... ñs, ms in old formations similarly > ns. Since nasals [n] were unvoiced before [s], the phonetic product was [ss] of all nasals + [s]. (OP1, PE19/47).

Some other combinations (a) occurred only in the earliest periods and were soon changed ... ns from n/m/ŋ + s. This soon became ss (with unvoicing of n) (OP2, PE19/82).

In Quenya ms was assimilated > ns (as mt > nt). The resultant ns, from any nasal + s, was then in the older strata unvoiced to ss (OP2, PE19/99).

The fact that these changes were in “old formations” and “occurred only in the earliest periods” indicates this phonetic development was in AQ or perhaps even earlier. There was a similar sound change in Old Sindarin, and in one place Tolkien said that ns > ss in Primitive Elvish:

s. This became [z] medially; but ns became prehistorically (common to Noldorin and Quenya) > ss. (Since later Q. tolerated ns this is sometimes found for pa.t. ss in late classical Q. and TQ.) (PE22/103).

As indicated above, the combination ns could appear in Tarquesta [TQ], and thus it seems this pair of letters might arise later, either from later compounds, from reformations by grammatical analogy or after [θ] became [s]. A note describing these reformations appears in OP2:

Reformation of this ss seldom occurred in PQ. They were more frequent in TQ especially in the past-tense forms of bases with original medial s. In both PQ and TQ they occurred only after the change of intervocalic s > z. The restored forms thus showed nz and were thus distinct from TQ ns < nþ with restored nasal before medial th. Since in Ñ[oldorin Quenya] s > z > r the nz forms were also obscured becoming, or being replaced by, rr (or rn). This nz remained in Vanyarin but (where not altered) became ns in Ñ[oldorin Quenya]. So borrowing from language of Atani inzil (Núm[enórean] inzil) “flower, lily”, became in Exilic Ñ[oldorin Quenya] insil (OP2, PE19/99).

Tolkien considered similar reformations in the 1930s:

At end of consonant groups [s] was stronger, and was not voiced: hense rs, ls, ns remained with (partly) unvoiced r, l {n}. The {classical >>} earliest product of ns was > ss; but ns was {occasionally >>} often reproduced by grammatical analogy in that case. But [later?] ns > nz as in Rothinzil (OP1, PE19/49).

Possible examples of past-tense formations involving primitive ns are:

• Q. hrinse, hrisse [second form unclear] past tense of [archaic] hríza < √SRIS “snow” (PE17/168).
• Q. virne strong past tense of virya- < √WIS “change, alter(nate)” (PE17/189).

In the first example it seems no reformation occurred (though since the present tense form is archaic hríza, maybe Tolkien simply didn’t write down the reformed past tense), but in the second example the ancient past tense may have been *winsē > *visse, reformed to virne. Other likely grammatical reformations include the reflexives insa “itself” and insë “herself, himself”, altered by analogy with forms like indë “yourselves” and intë¹ “themselves” (VT47/37).

Tolkien also seems to have considered having a later ns > ss sound change after [θ] became [s]:

ns > ns later Third Age ss: kanse, kāsen [with z written below kāsen] (in some notes associated with numbers from the late 1950s or early 1960s, PE17/95).

Where this grammatical renewal occurred in TQ after þ > s, ns remained as in TQ panse “smoothed” for PQ patte, after pase “smooths”. But panse soon > passe. This apparent substitution of ss for tt in cases of words having s for basic th is peculiar to Noldorin TQ (OP2, PE19/89).

For “substitution of ss for tt” compare also pitte past tense of ᴹQ. pise “spit” (PE22/103), probably with *pitʰnē, *patʰnē > pitte, †patte since [n] became [t] after aspirates. In the previous quote above, †patte was reformed to panse and then later became passe. If this later ns > ss sound change occurred, it was in late Exilic Quenya and may not have been reflected in spelling.

Finally, Tolkien also seems to have considered either later ns > nz (1930s, PE19/49) or nz > ns (1950s, PE19/99), but the only examples are tied to Ad. inzil “flower”, so its hard to say whether Tolkien adopted these ideas generally (see the relevant quotes above).

Conceptual Development: Based on the example ᴱQ. fansa “swoon” from the root ᴱ√FANA (QL/37), it seems unlikely this sound change applied to Early Qenya. Tolkien established the basic rule of ancient ns > ss in the 1930s and seems to have retained it thereafter, though as noted above he seems to have vacillated on the development of ns combinations that arose later in Parmaquesta.

Neo-Quenya: For the purpose of Neo-Quenya writing, I think it is best to assume that only ancient ns > ss and any later ns survived unchanged.