Paul Bartlett wrote:
>It seems to me that alleged"worldlangs" would differ little from a priori language such as Sona, in
that most learners would recognize so few of the roots that they would
be almost semantic wholes to be learned by brute force.

It depends. I'll give you a few examples that I have at hand to demonstrate that "worldlang" doesn't mean "unintelligible". From time to time they post a picture with a short text in Lidepla to the Facebook group of "Polyglots". Here are some reactions:

-"The language at the bottom looks like a mix between Chinese and something Germanic or Latin. Any clues? The gai means should in Chinese, and the vere means true in Latin languages and the shwo means speak in Chinese. It all makes sense to me, but is very strangely mixed."

-"That's pretty funny. I read the sentence at the bottom and was wondering why I understand almost everything and I freaked out for I couldn't guess what language it is until I was doing research on Lingwa De Planeta."

-"It sounds to me like Hindi. But it shouldn't. And that makes me laugh. :-/" ... "I must admit I'm rather baffled about the kind of expertise in each of these languages that seems to have gone into coming up with this conlang. 'Hi', used in this sentence ('se hi'), is a rather quirky grammatical particle in Hindi that literally means 'only', used to emphasize difference, uniqueness, or particularity; I've never seen a direct parallel in any non-Indic language.
That does imply that this was more than a menial exercise of looking up translations for keywords and stringing them together in a singular grammatical framework, there was actual knowledge of these '10 most spread languages' involved."

- "Well, I see that I can understand 95 % of it, when reading! (I haven't studied this language yet)"

Also, I remember one person claiming that he can understand *everything* in a Lidepla text without prior study. I am not sure that was true, but so he claimed. Anyway, that would be a rare exception, of course.