What do non-determinists think would be the testable or observable difference between a universe with and without free will?
I think it's important to highlight something here - history is fraught with people on both sides of "this is random" and "this is predictable in this way" being wrong. I would not be surprised to find that further research in physics over centuries ends up showing what we currently see as "randomness" is simply a very complicated system. Even if you are able to rest bigger arguments on top of quantum randomness, it's not exactly a sturdy base coming from the bleeding edge of physics.
It's also very common among in the community of this debate to get bogged down in the theoretical and not look back to the very real applications of free will concepts in the world. Many of the distinctions they make are not really useful for humans, and they haven't realized even the most basic parts of the arguments can have radical societal effects.
Looking at a regression model as an example in equation form:
Y = a + bX + e
Some "proportion" of a phenomenon is understood. However, many things are not immediately understood but only seen to be a factor in explaining a phenomena if looking at things from a very complete, long term perspective.
Something which we are unlikely be able to do.
Many people can feel when an action produces a ripple effect. Now think of tons of ripple effects interacting, some influential, some not, etc.
Anyway, back to the equation - we have +e the error term, which is the uncertainty that any such explanation will fully predict how something plays out.
I think what you're getting at is the error term also has characteristics.
One of those properties could be that certain things advance all, others are bad for all. Just to tie it back to your other comments.