When you win a government grant for say an experiment with 3 sequential goals and your hypothesis about the first goal was wrong, you literally just email the program contact the results and they’ll almost always allow you to change the next two goals in light of the new data. You are not paid to support only their preferred hypothesis.