The probability theory is just a religion that >99% of people believe in, just like the axiom of choice. An RNG can never be absolutely fair, but it's religiously believed be fair enough for practical purposes so that people agree to use it to play. It's your right to refuse to play with it, but hundreds of thousands of players have reached an implicit agreement to rely on it.
Probabilistically independent variables / random processes always have zero cross-correlation.
So what you're challenging is that causal independence doesn't imply probabilistic independence. That's actually up to human faith. E.g. if someone tosses a coin twice, the outcomes are believed to be probab. independent, but strictly speaking, they'll never be precisely independent (either causally or probab.) because the tosser's posture, mood etc. might depend slightly on how the first coin falls.
A similar argument applies to more sophisticated RNGs. They'll never be absolutely precise. However, they're subject to testing by special watchdogs which make sure that the key statistics* of the RNGs (like the observed cross-correlation) are close enough to fair ones so that every player can put sufficient trust in them.
* It's impossible to determine any probability distribution precisely from a finite number of observations. All that can be done is hypothesis testing. I.e. if, in 0.01% of all trials, the RNG is observed to do 'tricks' that an 'ideal' RNG would theoretically do in <0.0001% of cases, then the tested RNG is rejected as 'not close to ideal'.
There are many unreliable things in life. The probability of someone dying in an accident or of an accidentally contracted infection is alas non-zero, and it's believed to be way higher than that of a massively bankrolled winning player going broke because of the imperfection of a properly tested RNG. The threat of going broke undeservedly is miniscule in comparison with natural disasters and other inevitable threats. So it's best to stop worrying and focus on more beneficial self-improvements.