You're still implicitly assuming a time-constraint, though: the amount of time the person is willing to put into the problem before giving up. Presume someone puts in years of thought to a single lateral-thinking problem, and yes, they'll get it. Anyone will get it, if just by raw brute force, testing over every possibly combination of properties of the system that might have an underlying correlation. Most people just aren't willing to do that.

Given a finite amount of patience (or, equivalently, a time-constraint, which sets "patience" to a known quantity instead of allowing it to vary per individual), we can give a person an infinite stream of unit-sized lateral-thinking problems, and then see how many will be solved correctly before they "hit the wall." This is then a measure of their ability, in general, to recognize patterns quickly enough to put these insights to use: IQ.