The next big round of stuff out of our new machine learning/data infrastructure (and the ui improvements that go along with it) are going to have some nice stuff for classical listeners. You should give “randomized play” a fair shot. I’ve added more stuff to my library testing this feature in the past few months than the year prior. I think it will be a nice kick to get a lot of us out of our collector caves.
I’ve also been leading discussions about more tactical improvements recently, so that we can fit some of that into the schedule in the coming year too, for example: Grouping Issue Redocumented