First, blind testing is useful because it can eliminate the inputs from the senses not in question.

Your example of not being able to tell any difference between water and wine is a good one. In this case one is talking about what you can “taste” as opposed to “hear”.

So lets look at that, lets say by blind testing, you gradually diluted real red wine with tap water and you do this progressively until the people say there is no difference between it and water half of the time.
Now, at this dilution, the question is since there is no statistical correlation one way or the other, does this mean it is the same as tap water?
Also, what if your one of the people who can still tell there is wine present even though most can’t?

Statistically speaking, in the first question, it means that in that group, there was no actual correlation and so for that group (not individuals), there is statistically no “taste” difference.

Allowing the subjects to “see” the shade of the liquid would certainly produce a vastly higher positive because they could also “see” that the water is slightly “pink”, giving them another mental data point for the subjective judgement.
Going further, if they watched the dispensing process they would “know” even more about what they were tasting and only those daydreaming would get it wrong.

This part is what’s wrong about “not” blind testing, you include many other inputs other than just “what you hear” and one cannot remove that subconscious bias, one can only test “blind” where you cut off from those other inputs.

So what if your one of the people in that group that could reliably tell which was wine or water?
You have (statistically) a greater sensitivity than the group had, but if one continued to dilute it further, you would reach a dilution where you could not tell which was which.
At that point, even for you, there would then be no taste difference even though there were still X molecules of wine in the water.
On the other hand, if you were allowed to see the water’s color or you watched it being diluted and distributed, you could still “taste” which was wine.

Also, in medicine were big money is at stake, it is well known that “doing anything” can produce a placebo reaction. Taking a “sugar pill” that is thought to be helpful medicine causes a positive outcome between ¼ and 1/3 of the time. It is the act of “doing something” thought to have a specific outcome, that causes a false positive.
In the case of drug testing, progress was VERY slow until blind testing was used to compare the result of the test drug against the placebo while no one knew which was which.

In audio, big money or life and death is not the issue and in some ways getting to the bottom of things would not improve sales figures, so it is popular to cite the approach as flawed. I suppose it really is flawed too if sales is the measure of success and you make costly hifi trinkets

I would put it more like this.
IF you can’t hear any difference between X and Y with your ears alone, when not knowing which was which ahead of time (blind), what are you really “hearing” when the differences are then audible “with” prior knowledge (not blind)?

Your looking at the wine and watching it being dispensed that’s all..

Tom Danley