Good points, I'll address a few.

> Having a closed class of content words is a downright anomalous property for a _language_ to have. 
Languages where the set of verbs stems is closed do happen, in fact I posted the link to the paper on this mailing list recently.

Well, we have a few levels of closed-ness and in conlanging, even if you are naturalistic, they are difficult to avoid.

The Murfs are an imaginary people created by me. They live in an imaginary world where they coin words when they feel like it. The imaginary language has an open lexicon.

I sell the Murfs in plush form with a dictionary and license attached, which says I will sue anyone who coins a word. My legions of Murf fans have forums and mailing lists which shout down any attempt to coin a new word. The fandom lexicon is closed for legal and social reasons among fans.  The lexicon is open with respect to me because I can coin words when the whim strikes.

A thousand years pass, and I'm dead, all the original fans are dead. Now it is just a few scholar who study the Murf language as an artifact of 20th century culture.  Is the lexicon open or closed? Well, depends on who ever is using now, they no long have any social or legal restrictions. This sort of is like coining new words for sumerian. I think linguists would be reluctant to say that the newly coined words were part of the language and coining new words is on a the slope to a new language.  This sort of closedness is the fate of all conlangs and natural languages after they die.

Alternatively I could have created something that doesn't have imaginary people and isn't on a naturalistic design methodology, and the designer states: the set of content words is closed.  This sort of closedness has some interesting properties already mentioned and they are not interesting for their similarity to Elvish or human languages.

>I'm not conlanging for an audience so it doesn't especially rise to be a concern
I see this so often I need a name for it and I've been calling it the solipsistic methodology. In the solipsistic methodology, there really isn't anything outside of the conlanger's desires that matters, which makes asking questions on a mailing list a sort of a strange behavior.  I can only assume the OP asked the question because the answer has ramifications and a standard of truth that exists outside of his own mind.  The truth value properties about statements about solipsists language exist only in his own head so why should that person ask questions? The only answer in a solipsist methodology would ever would be, well, "How do you feel about it? Really? That is correct."  Anyhow, I encourage everyone to follow their joy, including solipsistic methodologies, I'm just a spectator here.

re: lexically closed vs derivationally open
This comes down to what one calls a word/lexeme or what have you. Even in toki pona, (and Greek and English!) there are phrases which behave like words but are clearly made up of phrases, like chemin-de-fer, jack-in-the-box, and the origin of goodbye, (god be with you), so even if the set of morphemes (bound and/or unbound) is closed, you can still make new things that pass a lot of the tests one might imagine for wordhood (e.g. does it take plural form, can it be split apart w/o changing meaning, does the location of accent change, etc)

re: language academy
Yes, this is a good idea, I think more conlang writers should include this sort of direction to fans because the lack of it is part of what prevents fandom in the first place. But to get to that, they will have to move past the solipsism, the "get-off-my-lawn", "get-out-of-my-sandbox" state of mind. But like I said, if that is what makes them happy, I don't want to be the one to try to push them out of that.

Matthew Martin