I've never heard of anyone saying "I thought that person was really intelligent, but they turned out not to be", and when there are scandals about people with fake credentials, they don't seem to come from people with fake credentials making mistakes-- instead, someone checks the history.
It seems to me that you can find out a lot about people's intelligence by talking with them a little, though I've underestimated people who were bright enough but didn't present as intellectual.
The real problems are with identifying conscientiousness, benevolence, and loyalty-- that's where the unpleasant surprises show up.
If this is true, then unstructured interviews should be a good way to determine how effective a candidate will be in a position. The literature is clear that unstructured interviews are worthless, and IQ testing is the best measure we have, typically explaining about half of the variance.
Lots of people have tried to dethrone IQ as a measure for a very long time, trying lots of things. They've never succeeded; IQ really is that good a measure of cognitive ability, and cognitive ability really is that important for almost everything.
Side note: my understanding is that the correlation is about one half, so about a quarter of the variance is explained.