Adjectives in Quenya generally (but not always) precede the noun they modify and generally (but not always) are inflected into the plural when they modify plural nouns. As Tolkien described it in Common Eldarin: Noun Structure from the early 1950s:

Adjectives were not distinguished inflexionally from nouns in Eldarin. But it is probable that the arrangements in Eldarin syntax were the same as in Quenya. Adjectives normally preceded the qualified noun (cf. the position of the qualifying noun-stem in “loose composition”), and in attributive use were seldom separated from it by other words or elements. (A standing exception was made by numerals which usually immediately followed the noun.) They in fact made “loose compounds” with the qualified noun, and only the qualified noun was inflected. In Quenya attributive adjectives are inflected for number only, if they precede their nouns. If they follow, the situation is reversed (PE21/77).

The main exception to this rule is that adjectives might follow nouns in poetry, as an “abnormal order only permitted in verse” (PE21/77). One reason why adjective word order is important in Quenya is because the copula ná- “to be” is optional. Thus “[a] man is angry” is usually written atan raiqua (less frequently atan raiqua ná). Thus raiqua atan “[an] angry man” and atan raiqua “[a] man is angry” can be distinguished from each other only by their word order.

An adjective following a noun, or if preceding separated from it (even by an article) was in Eldarin, and also in normal Quenya, predicative. So Q Sindar i Eldar Malariando "Grey are the Elves of Beleriand"; i rokkor rindi “the horses are swift”. In predicative use the adjective was inflected as the described noun (PE21/77-78).

Following Tolkien’s later pluralization rules for adjectives (see below) the first of these sentences would be sinde i Eldar Malaraindo rather than sindar. Adjectives are still inflected into the plural when used predicatively, as in: vanime Eldar “beautiful-(plural) Elf-(plural)” versus Eldar vanime “Elf-(plural) [are] beautiful-(plural)”. Things are a bit more obvious when you use the definite article: i raiqua atan “the angry man” vs. raiqua i atan “angry is the man”. That’s because the definite article i cannot intervene between an adjective and its noun.

Plural Adjectives: With rare exceptions, adjectives in Quenya end with either the vowels -a or -e or with the consonant -n. The last two form plurals like nouns, changing the final -e to an -i or adding an -i for adjectives ending in a consonant as in: alahendi “eyeless” plural of alahen (PE22/153) and yuluiti “waterbreathing” plural of yuluite (PE22/155).

By far the most common adjective ending is -a, however, and it has a plural ending unique to adjectives, -e: rembine “entangled” plural of rembina (PE17/26), néke “pale” plural of néka (MC/222-223). That is because Quenya adjectives did not adopt the noun plural -r, instead retaining the more ancient plural ending -ai which at the ending of words underwent the normal phonetic change to -e at some point during the Parmaquesta period (Classical Quenya). Tolkien mentioned this specialized adjective plural in a draft of a 1955 letter to David Masson:

final -ai in plural of adjectives > e, linte, unótime etc. (PE17/76).

The use of this specialized adjective plural is one way you can distinguish ordinary adjectives from adjectives used as nouns. For example, the ordinary plural of the adjective sinda “grey” would be sinde, but when used to describe the Grey Elves it uses the noun plural Sindar. It seems that nominalized adjectives are frequently pluralized as nouns in Quenya: a vanimar, vanimálion nostari “O beautiful [ones], parents of beautiful [children]” (LotR/981; Let/448). In English you might say “the Rich [singular] control our country” but in Quenya probably Herenyar [plural] mahtar nórelva.

Another common class of Quenya adjectives are those ending in -ëa, and those have yet another specialized plural ending: -ie: elvie “like stars” plural of elvea (MC/222-223), laurie “golden” plural of laurea (RGEO/62). This is again the result of normal phonetic developments: -eai > -ëe > -ie.

Adjectives and Noun Case Endings: Adjectives are not declined into a noun case in order to agree with a declined noun, as in sinda Eldo “of the grey Elf”, not **sindo Eldo (PE21/77). However, due to the “last declinable word” rule (as described in the entry on noun cases), an adjective might be declined into a noun case if it happens to be the last word in a noun phrase. This might happen if it is a personal epithet: Elendil Vorondo (genitive of Elendil Voronda “Elendil the faithful”; UT/305, 317). It can also happen if the adjective appears after the noun in abnormal/poetic word order: isilme ilcalasse (locative of isilme ilcala “moonlight gleaming”; MC/222). However, in verse word order is quite free, and the adjective might even appear after the declined noun, as in: ondolisse morne “on the dark rocks; (lit.) on rocks dark” (MC/222).

Roman Rausch looked at adjective agreement in some detail in his 2006 article on the Agreement of Adjectives in Quenya (AAQ). He noted that in Tolkien’s later writing, adjectives modifying nouns in plural noun cases are generally not declined into the plural:

• ilya raxellor “from all dangers” (VT44/5).
• aran linta ciryalion and aran linta ciryalíva “*king of swift ships” (PE17/147).
• atalantea mindonnar “upon fallen towers” (MC/222).
• ve quenderinwa koainen “as with bodies of Elvish-kind” (PE17/174).

In the second-to-last example, the first iteration of this phrase had a plural adjective, revised to non-plural: atalantie mindonnar >> atalantea mindonnar. In the last example, Tolkien modified the phrase from a non-plural to plural adjective: ve quenderinwa koainen >> ve quenderinwe koaron. Still, the bulk of the examples show no plural agreement with nouns in noun cases. This syntax was discussed all the way back in the Early Qenya Grammar of the 1920s, regarding adjective agreement for nouns declined into the adverbial cases:

where a qualified noun receives one of these endings (a somewhat archaic mode) the adjective usually precedes UNINFLECTED (except rarely for plural), and is virtually a loose compound, being often so written, as: TARA·KASSE “on the high head (or top)”; pl. TARA·KASISSE(N), or rarely TARE KASISSEN (PE14/78).

Assuming the same rule applied in later Quenya of the 1950s and 60s, plural agreement was allowed but rare when an adjective preceded a case-declined noun. Note, however, that when the adjective followed the declined noun (in poetic word ordering) it was always put into the plural to agree with the noun, though there are only a couple of examples: rámainen elvie [vs. singular elvea] “on wings like stars” and ondolisse morne [vs. singular morna] “on the dark rocks” (MC/222).

Rausch’s article also gave a number of examples where the expected plural agreement between adjectives and plural nouns did not occur even though the noun was not declined into any noun case. Those examples mostly date to the 1930s, 40s and early 50s, I would recommend ignoring them in favor of having plural agreement.

Conceptual Development: Agreement between plural adjectives and plural nouns dates all the way back to the first known Qenya poem, Narqelion, where the normal plural marker was adding -i to adjectives ending in -a as in ᴱQ. úmeai plural of úmea “large” and kuluvai plural of kuluva “golden” (VT40/8). There are, however, other plural adjectives with -r, as in ᴱQ. amaldar plural of (a)malda “soft”, and such r-plurals appear in the Lost Tales, as in ᴱQ. Tolli Kuruvar “Magic Isles” (LT1/85).

The basic plural adjective forms given in the Early Qenya Grammar (EQG) of the 1920s are remarkably similar to the ones appearing in Tolkien’s later writing:

Adjectives (except as above) agree in number and case with their nouns. They appear only in the forms -a (vocalic); -e, -o (rare), semivocalic; and consonantal (usually -n < nd). The latter are not as common as -a, -e, which form the bulk of adjectives ... Sg. anda, pl. ande (< ai) ... Sg. ninqe, pl. ninqi ... Sg. melin pl. melindi (PE14/45).

This is essentially the same as the Late Quenya system described above, except that adjectives both “agree in number and case with their nouns”. Tolkien also gave accusative, genitive and dative forms for adjectives, basically matching the corresponding noun inflections. This was not, however, true for the adverbial inflections like -sse, -llo, -nta (which would become -nna in Late Quenya), where such agreements were allowed only in poetry:

The adverbial suffixes ... they are never added except in verse to an adjective in agreement with its noun: where a qualified noun receives one of these endings the adjective usually precedes uninflected (unless for plural) and may be written as loose compound, as tára·kasse “on the high head” (PE14/46).

In the typescript version of EQG, Tolkien went further, saying that with adverbial suffixes adjectives were only rarely inflected into the plural to match their noun, usually tára·kasisse(n) rather than táre kasissen (PE14/78, see full quote above). As pointed out by Roman Rausch in AAQ, Qenya poems from the 1920s have a fair number of examples of adjectives and nouns being both declined into various noun cases, such as locative ailinisse oilimasse “shore-(locative) last-(locative)” (PE16/62), nominative ondoin morin “rock-(nominative)-(plural) dark-(nominative)-(plural)” (PE16/62) and nominative wingildin o silqelosseën “the foam-maidens with blossom-white hair” (MC/216). There are no signs of adjective-case agreement after the early 1930s, however.

Plural agreement for adjectives continued to be a feature of Quenya up until the 1960s, but Tolkien vacillated somewhat on the exact plural endings. In the 1930s up through the early 1950s, the usual plural ending for adjectives was -r: 1930s ᴹQ. írimar “beautiful-(plural)” (LR/72), 1940s ᴹQ. raikar “bent-(plural)” (SD/310), early 1950s Sindar “grey-(plural)” (PE21/77). By the mid-1950s, however, adjectives ending in -a consistently showed plurals in -e, and where they had r-plurals they function as nouns, such as naukar “the Stunted” referring to Dwarves as a people (PE17/45).

Neo-Quenya: Adjective agreement with plural nouns is the norm in Neo-Quenya, but outside of Rausch’s article I haven’t seen any Neo-Quenya courses suggesting that such agreement would not occur if the noun is declined into a noun case. Nevertheless, most of the examples from the 1950s and 60s suggest that tára orontissen “in the high mountains” would be the normal form, and táre orontissen would be rare. I would treat either as acceptable for Neo-Quenya.