PE19:101 helps; "Of zg: √MASAG: mazgō/ŭ 'sticky substance', Q makso 'mire'; Tel mascu. In Sindarin zd > 'd, đ, but zg, zb > đᵹ, đb > đa, đu as in nadha 'fetter' [Q naxa from √NASAG > nazg-], maða 'mud', buðu 'large fly.'

The PE19 fragment may remind one of John-Morris Jones' "A Welsh Grammar" § 97.iii., with W twddf ('a swelling') < *tuδʒ- < *tuzg-. ;-) (Said opus book being Tolkien's well-known recourse for Welsh.)

—

I think it's also conceivable the N/S verb(s) meaning "knead" might contain a prefix (like Italian impastare, Sp. amasar, & the various N/S verb formations we know).

I'd guess that ʒ vocalizes to a only word-finally, word-medially it should give i. Compare tara < *tarʒ, but tarias < *tarʒas (TARAG-) and similarly thala/thalion (STALAG-).
In fact, the original entry of MASAG- had madhias 'dough >> sofness, pliancy' (VT45:32).
So all in all I'd reconstruct *madhia- or *me(i)dhia-.

Thanks, Roman.
I've seen some  maða, buðu etc. "updated" in neo-Sindarin usage to mað, buð because Sindarin doesn't retain final vowels; given the comparisons to tara, thala etc. do you think that is incorrect? I'm admittedly quite rusty at Sindarin.

That makes no sense to me.. Tara & thala are strictly Noldorin, of course, but PE17:118 reiterates previously Noldorin fela with vocalized ʒ, PE17:148 has madu with syllabified w; and I cannot think of any example where the resulting vowel would be dropped in such a case...

Good to know! What do you think the plurals of nadha, budhu would be? nedh(a)i and bydhy? Would fela have feli?

Uff, good question.. Since the only thing attested is fela > fili with final-syllable umlaut (apparently it happens before vocalization), I would assume *naidhi and ?bydhi from *naðʒi and *buðƀi.
It looks like ʒ vocalizes to -i finally under influence of pl. -i (perhaps it became palatalized). This may be different for ƀ, it might become *y or *ui instead, hence ?bydhy or ?bydhui.
In any case, at later times by analogy we should have *nedhai and *bydhy.