...is that they always end up with inflated phonologies. What is the point of an international auxiliary language that takes learners years to learn to pronounce correctly?
I was going to say that, in the earlier languages, there just wasn't enough understanding yet about these things for their creators to make good choices; but Volapük actually did a surprisingly...acceptable job. In the consonantism, anyway. Well, at least Schleyer tried.
I'm not just talking about big, highly publicized projects, though I should probably mention in passing some of Esperanto's monstrosities, such as sciuro [stsi`uro] "squirrel." Virtually every...no, scratch that: every auxlanger seems determined to ignore the fact that vast chunks of the world population will be unable to distinguish between at least some of the phonemes of his purportedly ideal language.
Seriously, taking in view comparative phonologies, the inventory of an IAL should really never be any more complex than the following:
p t k Ø
Even such a small system requires a number of provisos, around palatalization (e.g. /s/ and /t/ before /i/), free variation (realization of /r/, particularly), for example.
(Note: There is, of course, a certain amount of wiggle room within the framework of a "maximum mininum" phonology. It's up to the creator's largely orthographic aesthetics, for instance, whether /r/ or /l/ is chosen as the base phoneme, as many languages have only one or the other and speakers will need to recognize it regardless of its position on the alveolar sonorant spectrum. /y/ and /w/ are quite common cross-linguistically, and might be considered for inclusion, but I've omitted them here because I would hate to see the distinction between e.g. "kiya" and "kia" bear a semantic load.)
Anything more than this and there's got to be quite a lot of willing suspension of disbelief going on right from the start. Of course, it's obvious why no language with a phonology like this has (to my knowledge) been submitted to the IAL-interested public: mathematics. With nine consonants and three vowels, we have a pittance of basic roots: 27 monosyllables, 729 bisyllables, 19,683 trisyllables. It's difficult to work with so little phonetic material.
Still, for those of us who do believe an invented IAL for the world would be cool, there it is. Either we should consciously say, "Well, working within real cross-linguistic phonetic limitations is too constricting, so I'm just going to pretend they don't exist, even though this means that this is mainly a form of artistic expression rather than a real attempt to create an IAL," or we should accept that every real attempt at a worldwide auxiliary language from here on out is going to have to wrestle with this problem, and probably have much longer words.
Maybe studying Rotokas should be a prerequisite.