Nothing will make you feel as humble as reading cutting edge linguistic papers about language features. If a language, say Cree, has a feature and no one in academia agrees on how it works, one can still go that community and learn Cree. On the other hand a constructed language’s learners have to rely on the specification to speak the language. If the meta-language (i.e. the language use to describe the language) is hard enough to read, then people won’t grasp how the feature works and won’t actually use the feature as designed. Either the fan community will work out a possibly similar but different system, or just fall back on the closest structure in their L1.
Now if we use old fashioned meta-language, then we get to fall into another trap of specifying a conlang as some subset of Latin or English, depending on what popular mental model of language you are using, being medieval or modern 5th period high school English class.
And finally, the linguists working with real language are looking for universals. They have an agenda and it will keep the grant money flowing for years. They also have a fix (and dwindling) stock of data, and the data is likely all descendant from proto-human, i.e. all the languages came from language before, either as a mother tongue or a hybrid contact language. So the metalanguage that linguists are using, is optimized for working with the materials available. If human’s ability to use languages isn’t so built in that we can’t learn new tricks, then the natural language’s metalanguage will be restrictive and seem to exclude certain features.
On the otherhand, if the natural language linguists are right (that languages systems are universal) then a conlang designer at best is only picking some parameters. Worst of all, the current metalanguage for describing these parameters is inaccessible.
Hobbyists, if only because they only have an hour in the evenings and a few hours on the weekends, have to be practical. I plan to use as much linguistic theory is possible to understand and write in a way readable for fans, be skeptical about what features a language must or must not have, and make up my own metalanguage when the linguistic communities metalanguage is unintelligible.