One nice thing about global state management (re "our custom C# Redux implementation"), is you can backtrack and replay altered input logs. So inputs with diver

One nice thing about global state management (re "our custom C# Redux implementation"), is you can backtrack and replay altered input logs. So inputs with diverse latencies (eg keys with speech or vision or gestures) can be combined (silly example: click item while saying "trash"). You can react promptly to a click action, and then backtrack when 1000 ms later you find out that was instead a click+"trash" action. Or when incremental tts updates from "launch the missiles" to "lunch is mussels, in butter".