Ekmar-Tenkar is, to some extent, an exercise in maximally free word order, accomplishing it by means of suffixaufnahme (another project, Kulaqil, approaches the same problem from the other direction: by extensive head-marking on verbs). Now I'm considering using both suffixaufnahme and surdéclinaison in this project.
Suffixaufnahme and surdéclinaison are two different phenomena that can result in nouns appearing with multiple case inflections attached. Suffixaufnahme (which means "suffix absorption" in German) is when nouns in cases that modify other nouns (like the genitive case) are inflected to agree with the nouns they modify, so e.g. for "The rabbit ate the farmer's carrots", "carrots" would be in the accusative case, while "the farmer" would be in the genitive and also have an affix agreeing with "carrots" in the accusative case (and possibly gender and number, or whatever categories nouns show in our hypothetical lang). Surdéclinaison (which means "super-declension" or "over-inflection") is when an already fully inflected word form can be treated as a stem and inflected for case again. So e.g. in a lang with only one case for noun adjuncts, if you wanted to specify "the carrots in the garden", you'd inflect "garden" for a locative case, but since locatives are a type of verb modifier here, you'd add a genitive case inflection as well to show that it's actually modifying a verb. Or, going the other way around, in "She's at John's", you'd put "John" in the genitive and then add the locative to show that it is the thing owned by John that is the location.
Suffixaufnahme isn't terribly rare: it's found in languages of the Caucasus, Australia, and the Middle East. Surdéclinaison, on the other hand, is attested only in Basque (though the second example I used of its use shows that we can do something similar in English, just not with case marking exactly). They're not known to appear together in any language, but since the sample size for the latter is so small I don't think that necessarily suggests that they're incompatible.
Obscure linguistic terminology lesson over. Now to talk about
I've decided that Ekmar-Tenkar will have a noun class system, with 5 classes: title, sapient, animate, inanimate, and intangible. These are inherent to noun roots, and mostly break down semantically, though there are some potential gotchas (for example, "fire" and "wind" are both animate, but "air" is intangible). The "title" class is an oddity: it's the class for personal pronouns, but also some honorific titles. In some cases there may be closely related words with homophonous stems but in different classes, but these are still considered separate words. Noun classes are declension categories, so they determine the forms of case/number affixes. This means that re-declining a genitive (the "John's" example) also assigns a class depending on the choice of affix. So if we take the fully declined word meaning "John's" John.SAP-GEN.SG and re-decline it, we can refer to various things owned or controlled by John: John's [person] (sapient), John's [animal, such as a pet] (animate), John's [object, concrete place**] (inanimate), John's [abstract place, action, or concept] (intangible). By re-declining a case form other than the genitive, we can also refer to things with other connections to John. Let's go back to the "She's at John's" example. If the word form we're re-declining is the genitive John.SAP-GEN.SG, that would refer to a place owned or controlled by John, but if it's associative ("with John") John.SAP-ASSOC.SG, it'd be a place that he is associated with (e.g. a place he frequents), or if it is itself locative John.SAP-LOC.SG, it'd be a place where he currently is at. If it's ablative, it'd be the place he came from, allative the place he's going to, and so on.
Speaking of locations, I had decided that the local case suffixes could take "location specifiers" that derived more specific cases. So a specifier meaning "top" when combined with the locative would mean "on", with the allative "onto" and the ablative "off of". This would be included in agreement. The idea was that an older form of the language would use a strategy like modern Japanese, where you'd express things like "onto" with a noun meaning "top" in the allative, modified by the main noun in the genitive (or maybe the noun-adjunct locative in this language?). Later they would start compounding these location nouns with the main noun (e.g. "table+top"), and this would still later be re-analysed as part of the case system and trigger agreement. Now I'm not so sure. I may drop the specifiers entirely and just use the Japanese strategy.
Ekmar-Tenkar nouns also inflect for state: topic, definite, construct, and indefinite. This is a bit like including the article in the noun itself. This doesn't take part in agreement. Construct state marks possessees (this is a well-known phenomenon in Semitic languages). I had a vague idea of having construct state marking interact with the state of the possessor, and/or be partly determined by whether the possession relationship is alienable or not, but I've never nailed it down. Indefinite would be marked with a null affix (or, if you prefer, unmarked nouns are indefinite while other states are marked).
All this has made me aware that I need to give a little more thought to the name of the language itself. My usual gloss*** has been "the speech of Ek", where "ten" is the stem meaning "speech" and "Ek" is a toponym. This I've broken down as Ek.INTANGIBLE-CONS-GEN.SG-NOM.SG speech.INTANGIBLE-DEF-NOM.SG, but now I realize I'm not sure where the state suffixes are, and where the genitive singular is hiding. It's easy to figure the "-k-" in "tenkar" as the definite suffix and "-ar" as the intangible nominative singular, but in that case "Ekmar" is harder to analyse: is the "-m-" in "Ekmar" the construct state marker, and if so where is the intangible genitive singular suffix****? Is construct state zero-marked instead of indefinite? Is there deletion going on somewhere? Is "Ekmar Tenkar" actually a fossilized phrase that doesn't fit the current grammatical rules? And what about Scarecrow's brain?
All this focus on nouns, and verbs have gotten pretty short shrift. I do know they inflect for pluractionality, but beyond that I haven't decided on much of anything. I have a vague sense that there will be sentence-final particles, at the very least a question particle (hí) and possibly ones for evidentiality and maybe mirativity. But on the other hand there are a lot of things about this language that resemble Japanese, and I don't want to wander too far into nihongoclone territory.
Not sure if there will even be a distinct class of adjectives, or if adjectival meanings will be handled entirely by nouns. I'm leaning against stative verbs.
Phonology (and othography) is funky and a topic for another day.
*Technically, he linked this post, where he talks about surdéclinaison and verbs, but it linked to the post on nouns as essential background, and that's what got me going.
**In Ekmar-Tenkar, places that correspond to a physically defined object, such as a house, are typically considered inanimates, while places that are defined politically or socially or that do not have well-defined or clearly visible borders are considered intangibles.
***The name actually predates the project: "Ekmartenkar" was the name of an earlier, otherwise unrelated conlang sketch that I abandoned, but I recycled the name because I liked the sound of it.
****And don't say "It's intangible!" :P