In the (Old) Noldorin of the 1930s, it seems that primitive [s] in combination with a nasals [m] or [n] developed into [mm] or [nn]. There are several examples in The Etymologies:

• ᴹ✶asmalindē > ON. ammalinde > N. emlin “yellow bird“ (Ety/SMAL).
• ᴹ✶kas-ma > kazma > ON. kama “helmet” (EtyAC/KAS).
• ᴹ✶besnō > ON. benno > N. benn “man, male” (Ety/BES).

Tolkien also alludes to the sound change sm, sn > mm, nn in his discussion of the Noldorin use of the Feanorian Alphabet from the 1930s:

In later Gondolic (after Y.S. 300) medial mb > mm. The letter t thus became used = [mm] of any origin (PQ pm, bm, sm, mm, mb) and is usually transcribed mm ... In post-Exilic, Toleressean, [mm] generally > [m] but this was not recognized usually in spelling ... In late Exilic (after Y.S. 307) [nd] > [nn]. The letter 5 thus became used of [nn] of any origin (PQ tn, dn, sn, nn, nd); and is usually transcribed nn, except in ancient historical Gondolic names: as Gondolin (PE22/35).

David Salo noted this development, theorizing the medial sound changes resembled the corresponding initial developments, with the s first unvoicing the nasal and then assimilating to it with later revoicing (GS/§4.70): [-sm-] > [-sm̥-] > [-m̥m̥-] > [-mm-]. The above example ᴹ✶kas-ma > kazma > ON. kama “helmet” (EtyAC/KAS) hints at another path for the medial development, with the s voicing to z and then assimilating to the voiced nasal: [-sm-] > [-zm-] > [-mm-]. Tolkien mentioned such voicings of s did not happen in Primitive Elvish but were possible in Eldarin child languages, though he did not say specifically whether it occurred in Noldorin:

s became voiced whenever, by suffixion or abnormal vocalization of bases (see above), it came to stand before a voiced stop: thus sb, sd, sg > zb, zd, zg. Of these, only zd could be produced by suffixion; s was not voiced before other voiced consonants (as m, n, l, r, w, y) in Quendian or [Common] Eldarin, though voicing occurred in such positions in some of the individual languages (first version of Tengwesta Qenderinwa [TQ1], 1930s, PE18/54).

Similar sound changes happened in Welsh, where such s-assimilation occurred before both nasals and liquids: -sm-, -sn-, -sl-, -sr- > -mm-, -nn-, -ll-, -rr- (WGHC/§95ii). There is one example that hints that in Old Noldorin, -sr- might also become -rr-:

• ᴹ✶gaisrā > gǣsra > ON. gērrha > N. gaer “dreadful” (Ety/GÁYAS).

This does not seem to be the case in Sindarin, however:

• ✶cas-raya > S. cathrae “tressure” (VT42/12).

There are similar examples where [-sr-] > [-θr-] in Noldorin:

• ᴹ✶la(n)sro-ndo > N. lhathron “listener” (Ety/LAS²).
• ᴹ√LUS > N. lothren “wild, waste” (EtyAC/LUS).
• ᴹ√OS > N. othrond “fortress, city in underground caves” (Ety/OS).

For lhathron, it seems the consonant cluster was [-nsr-] in the Old Noldorin period, and only became [-sr-] later after nasals vanished before spirantal clusters. The Old Noldorin cluster was probably [-str-] in the other two examples: *lost-ren > *lhosren > lhothren and *ost-rond > *osrond > othrond. Thus, it’s possible that [-sr-] > [-r̥r̥-] > [-rr-] was the Old Noldorin sound change and [-sr-] > [-θr-] was the later Noldorin sound change after consonants were lost in larger clusters.

Conceptual Development: There are not many attested examples of these sound combinations in earlier iterations of the language, but it seems that in Gnomish at least the medial [sm] had the same development as initially, to [f]:

• ✶Eðusmānī > G. Edhofon (GL/31).
• ✶smaika- > G. faig “cruel” (GL/33).

As noted above, it seems unlikely that medial sr > rr in Old Sindarin, but it is possible that sm, sn > mm, nn, since a similar change happened in Welsh. There is no clear evidence either way.