There were a bunch of minor versions between 3.0 and 4.0 made up of fundamental refactor work to establish separation of concerns. As I understand it from the call graphs he shared, the problem was that the UI was reaching deep into the audio code directly instead of using a clear API boundary to separate concerns and route commands and change state. So fundamentally while the UI worked, it also was deeply entwined with the audio engine and changing any of the audio back-end basically broke things downstream in unexpected places and couldn't be swapped without deep UI work anyway. So, if you're going to have to rewrite your UI layer to disentangle it from your back-end anyway why not also bring it up to speed? Let me repeat that to be clearer: The UI had to be changed internally in fundamental ways to do any work on the audio back-end and they chose to also uplift the design and UX while making it easier to replace the audio engine in a future commit.