IANALinguist so take whatever jargon with a grain of salt.
I think there are two parts of conlanging that I really enjoy: the social-political aspect and the semantics of the languages themselves.
I was introduced to conlanging through toki pona & Esperanto, both of which are languages that take a strong political-philosophical stance on unity, simplicity, cross-understanding, etc etc. The collaborative meaning-making that tp requires is one of the things that draws me to it day after day. I was thinking about this re: on fluency in toki pona - #15 by TheAndSys this post. The conclusion that I draw here is that tp cannot take a concrete form without its community – the speaking of the language simultaneously builds it, and thus the language reflects the community that speaks it. The community is the thing, so to speak. (Also makes me think of Viossa)
This lack of a governing body on toki pona + the existence of reference materials that partially lexicalize the language (lipu pu, lipu ku, Tatoeba, whatever else) + the fragmentation of the community creates a situation similar to Arabic imo. You have a “standard Arabic” spoken in cross-regional news programs and with unfamiliar people, and then you have regional languages (i.e. Egyptian Arabic, etc) that are spoken in more familiar settings. Toki pona has taken on a similar sociopolitical fragmentation, splitting into multiple closely related languages sharing a common near-ancestor from nasin pi lipu pu. I absolutely love the politics here. It fascinates me to no end (& because of this I eat up like everything @gregdan3 writes lol). As far as design priorities go, I suppose this can only happen with a language willing to give up its ownership into the community’s hands. Contrast this with Lojban’s many reforms, which are nowhere near as pervasive or widely accepted, historically because of the LLG’s grip on the language and today because of many notable Lojbanists’ feelings towards CLL purism.
As far as semantics go, I feel similarly to @samflir in that I think it’s very cool when a conlang takes on some unique, idealistic way of meaning-building that differs from how a natlang might usually do it. On this front, I think about toki pona’s semantic subsets (see [2/100] toki pona and the semantic subset | DK’S ABODE and semantic spaces dictionary), Toaq’s serial predicates (Toaq - Syntax), or Eberban/Lojban’s slots & the existential claims you can make by not filling them. Lojban specifically has a ton of tools for meaning-making that throw me for a loop, like quantification through prenexes and PA4 or tacking things into slots in odd places (like through be or modals). When I first started learning Lojban, the idea that lo ... ku rips meaning out from the first slot immediately got me hooked and opened my mind to what’s possible with semantics.