I personally feel like a deeper understanding of Bayesian probability has mainly just helped me to formalize things that are already obvious (the goal being to replicate what is obvious to humans in a computer, e.g. computer vision, robotics, AI, etc.). There have been few instances where it has actually helped me weigh evidence more effectively. But maybe I am missing some set of practical techniques.

Also, I was unable to parse the final paragraph that you wrote, would you mind re-stating it?