If your target level of accuracy is 100%, there is probably no medical test that can show anything.

Did you know there is no conclusive test for Alzheimer’s, IBS, migraines, and dozens more physical conditions, nor for any psychiatric condition? Do you think these aren’t real, or that they cannot be discerned to a useful degree of accuracy?

Well that's why when the law is concerned, it usually defers identification to a trained professional (e.g. as determined by a doctor ).

Besides that, there are things you can determine with certainty: The presence of a substance in blood for example

A physical examination by a doctor, possibly augmented by imaging, is an excellent way to determine sex if previous observations and tests have been ambiguous. This is rarely a great difficulty, but when it is, it's how DSDs get diagnosed.

So which tests with which results would you rely on?

Which bit are you having difficulty understanding? Make trivial observations if you want moderate accuracy (~99%), use a cheek swab or genetic test if you want greater accuracy, consult a doctor if you are still unsure.

Which part of my question are you having difficulty understanding? If this is so important and accurate surely you can name and explain the important and accurate tests, and their results. Maybe it isn't as simple as you say?

Why should the simplicity of the definition (which, by the way, is what sex means in mammals generally, not just in humans) mean that the testing has to be simple in 100% of cases? I’ve already described cases where it isn’t. That doesn’t change the definition.

Do you think natural phenomena don’t exist or can’t be useful unless there is a simple test to grant you the details?

I think there are tests that intelligent people can describe which demonstrate the phenomena exists. That's what science is.

I think my answer was larger than the box you drew for it, and now you’ve sealioned yourself into a very weird corner where you seem to think the existence of male and female developmental pathways is somehow negated by edge cases.