They would be so correlated even if they were a bad metric. Imagine a score that multiplies academic success and the highest Tetris score obtained within 100 miles of your birthplace, then subtracts 50 points if you're a girl. That would also be positively correlated with academic success.
IQ tests measure something, yes, but they're not magic. (See also: https://en.wikipedia.org/wiki/Goodhart%27s_law) I received a high-quality education including what amounts to IQ-test tuition (to pass an entrance exam); do you know of any studies that attempt to factor out this particular confounder?
Ah, yes. The g factor: “that which is measured by IQ tests”.
The main arguments for IQ tests, as far as I understand, that they correlate with an assortment of cognitive metrics, which all correlate with each other. This is a question. “It's the g factor” is not an answer; you've just stuck a circularly-defined label on the question (or, rather, on a particular frequentist analysis of the question) and declared it solved.
Do you see a problem there?
The g factor—or more descriptively—the theory of general intelligence has little support outside of the field of psychometrics. And the field of psychometrics is not known for it’s scientific rigor. Needless to say, there is little reason to believe that the theory of general intelligence has any grounds in reality, nor that the g factor exists outside of an operationally defined concept that might not hold any descriptive power for human behavior outside of a narrow setting of pen and paper tests.
It seems stable because of how age is adjusting for and the huge sweet spot where 64.2% of the population is grouped between 85 and 115. Basically people with significant developmental disabilities at 5 don’t suddenly become normal, but an IQ of 180 at 5 isn’t stable long term.
My point is this, IQ seems to have signal. It probably doesn't hold up in extreme circumstances. Psychology in particular seems to have a difficult time holding large enough studies to find universal constants that we should accept that IQ might be the best local approxination for now.
You can read papers saying IQ texts on 2 year olds are strongly correlated with adult IQ, but read the details and individual people’s scores varied wildly.
So sure in aggregate it seems consistent, but that’s largely an artifact of test design. Someone with mental handicaps so profound as to be non communicative will consistently score very low, it doesn’t mean much for the general population.
No serious IQ test claim to measure much more than 2 standard deviations up and down (meaning they all break down somewhere above 130-135 or below 70). I have done WISC/WAIS three times (as a part of a psychological assessment) and the conversion to IQ scores were within 3 points (the WAIS/WISC scale measures many different abilities, and can be converted to an "aggregate" IQ score). I asked about this and when people are healthy the tester said people tend to score very similarly on tests, even many years apart.
The issue with the underlying raw score at 90+ is it doesn’t map very well to a bell curve. Many people are quite healthy but a significant number are drastically reduced from their younger self.
Had it been completely pop-pseudoscience it would not be used by, say, the Swedish school system as a tool to gauge whether someone has special needs that warrant another form of school with extra support. The tests used are however not the same as most people have done online, like WAIS (but for kids. Can't remember the name of it).