Not sure what you mean by several. I took 4 (5?) over a 30 year period, and the results were all within a couple percentage points of each other.
However, my point was that small variations are not significant. Also, the correlation between IQ scores and intelligence in general is highly debatable - and that's what people are arguing about.
Also, just because one can game the test, or make its result less accurate, it doesn't mean the population level data is much less accurate. I'm reminded of the Body Mass Index - men, by strength training, can put on lots of muscle mass and become classified as "overweight" or even (rarely) "obese" without being fat. But the fact is that most "overweight" men are in fact fat, and not muscular. The fact that a test for an individual can have reduced or zero validity, does not necessarily threaten the validity for the overall population.
If everyone were to start gaming the tests, things might be different. Then again, being able to game the test might be a highly IQ dependent task, so the IQ test remains an IQ test.
Individual IQ cannot be precisely measured. However heritability and economics studies show that it can be measured with good accuracy.
1) small variations in IQ tests don't matter, like for example the difference between 120 and 140 being negligible and depends a lot on external factors, like rest, vitamins in your blood-stream and mood ... give a person several IQ tests over a period of time and he'll score differently every single time
2) you can increase your IQ score with a certain amount of points just by making your brain work harder. Physical exercises work too (professional athletes are above average)
Ultimately the most common complaint is that intelligence (while hereditary) is not something you can accurately measure and that IQ tests are inherently not reliable.