Tell me when was the last time a CS theory was proven false or needed to be adjust because of it did not agree with experimental data.
That happens a lot in physics and other sciences. That's the way how science make progress but no CS. CS is mostly maths. You provre something is correct in CS because it is consistent with a set of axioms. We never run any experiment to verify axioms.
A scientific theory can be proven wrong by an experiment. It is a the best approximation of the reality we know until it is disproven. On the contrary CS/maths lacks this.
Many Algorithms can prove false and need to be adjust because do not agree with experimental data.
Can you cite an example? I have read many algorithm research papers in the past. Typically there is a proof of correctness. It is like a proof in mathematics. You prove that the algorithm works based on some assumptions or axioms. i.e., you can derive the proof from a finite sequence of logical statements from the axioms. Then there it is usually accompanied by a time or space complexity analysis, which is also maths. If you can can prove that it is O(n * log n) time correctly, then there is no way to disprove it later.
If you propose an algorithm and claims that it 'works' based on limited experimental data, then yeah it works like science but this is not how algorithm research works.
Let's throw this shit over the fence and see if it sticks. If not, change it a little and try again.