Science answers very limited questions.
Science is not in the business of answering "why". If "science" claims to answer "why", it is most likely not science! "Why" answers are the realm of religion.
Science's main answer is to the question of "what". "What is the acceleration of a falling object?" "What is the effect of sunlight on a leaf?"
If science answers "how", then the answer is limited to inside the world of a scientific model. A model is a simplification of the real world and therefore it is an imperfect representation of the real world. In fact I think we can question whether the model's mechanisms are reality. Even though the model might make an excellent simulation of the real world, does the model actually describe "how" the real world does it?
Some models are very good. So good in fact that today we cannot measure the error between them and reality. Tomorrow we may measure the error and that may become a problem. The relativity model was developed after we saw errors in the Newtonian model. Today we have errors like the Pioneer anomaly and quantized redshifts that appear to show defects in current relativity-derived models.
How is it possible to have these errors? The answer is simple - all models are limited and have errors. This is the normal course of science. And ideally science will get right on it and come up with improved models that make the measurable error disappear.
UPDATE: The Pioneer Anomaly appears to have been solved. Heat radiating from the spacecraft pushes on it and has been slowing it down.