This root for “(morally) good” and “holy” things dates back to Tolkien’s earliest versions of Elvish, probably due to its long-standing connection to the name Q. Manwë, one of the most stable names in Tolkien’s Legendarium. The unglossed root ᴱ√MANA appeared in the Qenya Lexicon of the 1910s with derivatives like ᴱQ. mane “good (moral)” and ᴱQ. manimo “holy soul” (QL/58). Derivatives like G. mani “good (of men and character only), holy” and G. manos “spirit that has gone to the Valar” also appear in the contemporaneous Gnomish Lexicon (GL/56).
In The Etymologies of the 1930s ᴹ√MAN “holy spirit” appeared with derivatives like ᴹQ. manu/N. mân “departed spirit” (Ety/MAN). Earlier versions of the entry had the gloss “holy” (EtyAC/MAN), and an earlier version of the entry for ᴹ√MBAD has MAN- “blessed” (EtyAC/MBAD).
The senses “good, blessed, holy” were retained in Tolkien’s later writings, though sometimes the root was given in its augmented form √AMAN. In Quenya Notes (QN) from 1957, √MAN was contrasted with √ARA which also meant “good”, but with the nuance of one specimen that is “good of its kind” and hence “excels, without necessarily implying that others are bad or marred” (PE17/147). Elsewhere in QN Tolkien elaborated on the meaning of √MAN in more detail:
> √MAN “good”. This implies that a person/thing is (relatively or absolutely) “unmarred”: that is in Elvish thought unaffected by the disorders introduced into Arda by Morgoth: and therefore is true to its nature & function. If applied to mind/spirit it is more or less equivalent to morally good; but applied to bodies it naturally refers to health and to absence of distortions, damages, blemishes, &c (PE17/162).
In Definitive Linguistic Notes (DLN) from 1959, √AMAN “good (morally), holy, blessed, free from evil” was contrasted with √AYA(N) “treat with awe/reverence” and √MAGA “good (physically)” (PE17/149). In The Shibboleth of Fëanor from 1968, Tolkien said the root meant “blessed, holy” and was adapted from Valarin (PM/357 note #18), which is consistent with the fact that its derivatives were almost entirely limited to Quenya and not Sindarin; where derivatives do appear in Sindarin, such as S. Avon the equivalent of Q. Aman (PE17/162), they were probably loan words from Quenya.