What bothers me about seeking changes is that it looks like patching pieces as they fail, to make transitions work. And there is no way to actually tell what is correct. The way I did it initially for other states was to create a full stub branch of source->transform->sink, let it run on empty samples, and then issue more commands to see what needs to happen. I don't see how this could be reviewed without doing something similar, especially if it's difficult to reproduce with a proper test.
How did you test individual changes, apart from testing the game with the whole set?