Researchers in genomics are aware of this and that is why researchers are expected to correct their p-values for multiple testing, for example using q-values.
No that isn't the point. These variables are often associated with an outcome or phenotype in a statistically rigorous manner. The point is that knowing this doesn't get us anywhere.
Get us anywhere? That is research, small pieces to the larger puzzle. Biology is complex, a lot of molecules doing different things. We don't have perfect tools to study those molecules or their interactions. Genomics and statistics offer one path to their study, and they complement already existing methods.
I don't understand what you are trying to say. My point is also that the tools are imperfect, particularly when it comes to generating useful insights, as opposed to just correlations.
If your tools don't generate useful insights, then you are not using them right. A properly designed study, employing high-dimensional genomics data, can for sure generate mechanistic insights, not just correlations. It's all about your study design.
Some tools will never give you useful insights because they have inherent limitations, or the model being studied doesn't lend itself to an informative study design. Results get published anyway. Which is fine. My point was that many high throughput studies do not materially contribute to the advancement of knowledge. This is entirely compatible with your point that well designed high throughput studies can contribute.