The earliest iteration of this root was ᴱ√ẎATA “join” in the Qenya Lexicon of the 1910s with derivatives like ᴱQ. yarta “yoke” and ᴱQ. yatta “neck; isthmus” (QL/105). In the contemporaneous Gnomish Lexicon it had derivatives like G. gada- “join, connect, unite” and G. gath⁽⁾ “neck” < ᴱ✶yatt- (GL/36). It reappeared as ᴹ√YAT “join” in The Etymologies of the 1930s with derivatives like ᴹQ. yanta/N. iant “yoke” and ᴹQ. yanwe/N. ianw “bridge” < ᴹ✶yatmā (Ety/YAT). Above ᴹ√YAT there was a deleted root ᴹ√YATH with a single derivative ᴹQ. yatta “neck, isthmus” that was not deleted (EtyAC/YAK). ᴹQ. yatta may have been transferred to ᴹ√YAT, though as pointed out by Carl Hostetter and Patrick Wynne it might also have been transferred to ᴹ√YAK, which included among its derivatives ᴹQ. yat (yaht-) “neck” (Ety/YAK).
In later writings, Q. yanta and S. iant were used for “bridge” (LotR/1123; SD/129). The word Q. yanwë reappeared with the gloss “joining” in notes from 1969, initially derived from ✶yadme but this was deleted and replaced by √YAN (VT49/45-56). Carl Hostetter suggested that rejected ✶yadme probably reflected the earlier phonological developments seen in The Etymologies whereby tm became dm, then nm, and ultimately nw. However, sometime in the late 1950s or early 1960s Tolkien abandoned tm > dm > nm > nw, deciding instead that tm > tw, which may have motivated him to change the root ᴹ√YAT to √YAN. A similar revision can be seen in change of ᴹ√TEK to √TEÑ in the derivation of Q. tengwa “letter”.
Neo-Eldarin: For purposes of Neo-Eldarin, I prefer to retain the phonological developments of the 1930s to 1950s whereby km, tm > gm, dm > ñm, nm > ngw, nw. As such, I think it is better to retain the 1930s form of the root ᴹ√YAT “join”, which also helps avoid conflict with √YAN “wide, extensive, large”. As the above discussion indicates, Tolkien seems to have been vacillating on tm > tw vs. tm > nw as late as 1969.