After the voicing of s to z, the combinations zn, zm became nn, mm. This change is mentioned in both the Outline of Phonetic Development (OP1) from the 1940s and the Outline of Phonology (OP2) from the 1950s:

The earliest of these voicings, possibly already in CE, or at least AQ, was in sm, sn > zm, zn. The product of these in PQ was mm, nn. When later but before the general medial voicing of s the suffixes m, n were re-added to voiceless [s] the nasal was probably unvoiced and sm, sn > sw, st [per note #109 Tolkien wrote an “X” in the margin next to this and then “zm, nn”]. But cases are rare, and may be usually explained by suffix-alternation (OP1: PE19/49).

The change of sm, sn > zm, zn was the earliest. Here the phonetic change was > mm, nn. These were preserved in PQ and Noldorin TQ. In Ñ. after the change of z > r, in that dialect the etymological or grammatical relations of r to mb/nd became obscure and rm, rn were sometimes produced by re-formation as in √LAS “listen”, AQ lammo “listener” (spy or scout), Ñ larmo (OP2: PE19/101).

It is not entirely clear when zn, zm became nn, mm, but it must have been before [z] became [r]. The example AQ. lammo above indicates the sound change occurred in AQ; this word is also the best example of zm > mm. The best example of zn > nn is ᴹ✶besnō > ᴹQ. venno “husband” in The Etymologies (Ety/BES).

Conceptual Development: In the Qenya Phonology from the 1910s Tolkien said that s before n, m ultimately became rm, rn (PE12/19). Examples include:

• ᴱ√HOSO > ᴱQ. horma “horde” (QL/41).
• ᴱ√NESE > ᴱQ. nermi “field spirit” (QL/66).

The small number of examples makes it difficult to determine when Tolkien adopted the rules described in OP1 and OP2.