My only point is that there is no way to say "events with probability < X are impossible in real life", for X > 0. For any probability value, it's trivial to construct an experiment that will be guaranteed to have an outcome whose probability is that low.
My own view is that there something deep about probability akin to the measurement problem. My view is that probability is not an objective fixed thing, but that it is something relative to the observer of experiments.
This is basically the Bayesian view of probability - that probability is a measure of the knowledge of an agent, not a property of a system. There are even some interpretations of QM that try to find a way to apply this to the measurement problem.
However, for this particular case, I don't think there is anything all that mysterious. If a process is more or less equally likely to produce any of a huge number of outcomes, it stands to reason that you can't predict which outcome will happen, even while knowing some outcome is fully guaranteed.
I also know that it is impossible.
If you disagree, show some proof. Not "a proof". Proof.
I live in the real, not theoretical world.
Create a video showing one million dice rolls that have all hit six, and I will accept that it is not impossible to roll a die one million times and have every roll hit six.
No numbers. No computer programs. No math. Dice.
Mother. Fucking. Dice. Cubes covered in pips.
Real, physical, standard layman's definition dice rolled in the real, physical, actual world.
Show me! Please!
Until then? S'ay imposseebluh.