Anyway, it was enough to prime me to the idea that IQ might not, for an individual, be as stable and repeatable as all that. Since then, while I haven't made a careful study of the research, I've seen nothing to convince me that the probability of getting a number which is 10 or 15 points away from what it would be on another day is low enough to make it worth having a number that might be misleading in that way. I'm quite prepared to believe that for the purposes it's intended for IQ is quite good - but honestly, it isn't intended to distinguish 135 from 125 or 145, and I know I'd have a hard time not imbuing a number with that significance that it shouldn't have. So, better to look at the child.
The real problem seems to be that we need to figure out developmental arc over a lifetime, which I.Q. tests apparently can't do very well.
I suspect that such arcs are reasonably fixed, with some wiggle room, but not much.
The significance is the nature of the arc for the individual, not the score on a test on a particular day.