Let's say I constructed a language L and taught it to a thousand people. In a few years, their level of proficiency is perfect. I ask them to only speak L from now on and send them to an island where they make lots of babies.
Question: will their grown-up children speak exactly according to the way I originally constructed the language, no matter how hard it actually is?
Almost certainly not, because the rules of language change still apply. Without considerable engineering (whose methods have probably not been invented yet, or at least not been tested on a large scale and refined), it is likely that there will be a lot of corners that are subject to relatively quick change when people start to use the language.
Similar situations occur in practice which demonstrate this in the variant "no matter how easy the original language actually is". In certain contact situations, a very simple pidgin can develop and become so successful that many children grow up with it as their native language. Within a few generations it becomes a full creole language.
There is no reason to assume that one wouldn't see similar instability issues in the situation you describe, where the new language is complex rather than simple. (Or with a new language of average simplicity.) It's important to understand that the evolution of languages doesn't converge towards a regular, logical and practical state but is caused by factors such as people's desire for succinctness, vivid expression, politeness etc.
Note that complications in the original language are definitely not the issue in the way you seem to be worried about them. If the language is easy enough for the first generation to learn perfectly, then it's definitely easy enough for their children to pick up.
I agree with jlawler's comment that the hypothetical scenario is impossible. However, for what it's worth, a somewhat comparable real-life parallel is the case of Hebrew, which was "revived" after centuries of near disuse. Modern Hebrew has undergone a very considerable amount of change in the century or so -- mostly in vocabulary, syntax and idiom, but also in morphology, phonology and phonetics. The extent of change is greater than what any long-established language I know of has undergone in the same period. This suggests that no matter how carefully a constructed language is designed, once it is used in the real world it will quickly start looking very different from what the designer(s) had in mind. This is inevitable given normal human linguistic creativity, the inexorability of language change and the fact that no designer can anticipate all the uses a language will be put to. Since a lot of language change happens during childhood language acquisition, even one generation should be enough to bring about some change -- for example, if the morphology is complex ("no matter how hard it actually is"), you could expect to see some amount of analogical simplification.
