In Sindarin and Noldorin, [z] disappeared before the voiced stop [d], modifying the preceding vowel, though the exact phonetic rules evolved between the conceptual stages of the language. For Sindarin, this change is described in a couple places within notes on the Outline of Phonology (OP2) from the early 1950s:

In Sindarin on the other hand zd > d with lengthening of preceding vowel so early that these newer long vowels followed the same development as the original long vowels (ā > ǭ/au; ē, ī > ī; ō, ū > ū). The intervocalic d then > đ (dh) as usual (PE19/91).

In Sindarin zd > ´d, đ [with nearby example mizdē > míð] (PE19/101).

As these notes indicate, this sound change was quite early because the development of the resulting long vowels matched that of primitive long vowels. Thus, the change occurred no later than Old Sindarin and possibly even in Ancient Telerin, since the Telerin phonetic developments were similar. Compare ✶ezdē “rest” > S. Îdh vs. T. Éde, the Sindarin and Telerin names of Q. Estë (Quendi and Eldar, circa 1960, WJ/403). David Salo attributed this change to Ancient Telerin in Gateway to Sindarin (GS/§4.27), but he actually described the Noldorin sound change (see below), since the information about Sindarin developments had not yet been published.

There are a fair number of Sindarin examples of this sound change:

• ✶ezdē [> *ēde > *īde] > S. îdh “rest” (PE19/91, WJ/403).
• ✶mizdē [> *mīde] > S. mîdh “dew” (PE19/101).
• ✶reddā > rezdā [> *rēda > *rīda] > S. rîdh “sown field” (PE19/91).

The combination [zd] could arise in two ways: when primitive [s] became [z] before voiced stops (as in the first and second example above) and when primitive [d+d] from suffixion became [zd] (as in the third example above). Either way, the ultimate phonetic developments of [zd] in Sindarin were the same.

During the Sindarin conceptual stage from 1950s forward, it seems that combinations of [z] with other voiced stops developed differently, when [zb], [zg] became [ðβ], [ðɣ]. This was probably a much later phonetic development, around the same time that voiced stops became spirants more generally.

Conceptual Development: There isn’t enough information from the 1910s and 1920s for us to determine how [z] developed in combination with voiced stops in the Gnomish and Early Noldorin periods. By the 1930s, however, there are enough examples in The Etymologies to deduce similar patterns for Noldorin. For example:

• ᴹ✶ezdē > Eide > ON. Īde > N. Îdh “rest” (Ety/EZDĒ).
• ᴹ✶mizdē > N. mídh “dew” (Ety/MIZD).
• ᴹ✶mazgā > ON. maiga > N. moe “soft, pliant” (Ety/MASAG).

Although the end results were the same as the Sindarin development in a few cases, the first example shows that in Noldorin [Vzd] > [Vid] (instead of Sindarin [V̄d]). The second example shows that [z] vocalized before other stops as well, with [Vzg] > [Vig] (instead of Sindarin [Vðɣ]). There are no examples of [Vbd] in The Etymologies, but notes from Tolkien’s discussion of the Feanorian Alphabet from the late 1930s make the Noldorin development clear:

The sound [z] only occurred in zd, zg and rare zb. For each of these ON used a single letter @ S and W. The phonetic changes that overtook these groups do not therefore normally find expression in ON. The exilic development suggests that @ W S became during ON {[ð, ðb, ðg] >>} [i̯d, u̯b, i̯g] but these letters continued in use (PE22/26 and note #78).

This suggests that [Vzd] and [Vzg] > [Vid] and [Vig], while [Vzb] > [Vub], which is consistent with the examples in The Etymologies. However, an earlier version of this note (before being modified to the above) instead reads:

The exilic development (to ðw, ði̯) suggests that W S became during ON [ðb, ðg], but these letters continued to be used. zd, however, became d, with lengthening of preceding vowel, and in consequence in late ON 2 was often substituted for @. (PE22/26, before the editing in note #78)

These pre-revision sentences match the Sindarin phonetic developments described above. It seems that Tolkien was already vacillating between the Noldorin vocalizations and the Sindarin sound changes as early as the late 1930s, and perhaps continued to do so for some time afterwards, before settling on the Sindarin sound changes before or around 1950 (he may have vacillated after 1950 as well, but there is no evidence he did in the currently published materials).

See OS. [z] vanished before [d] lengthening preceding vowel for discussion.