See this clip: Desktop 2024.01.14 - 00.18.07.01
You can see the IO starts of supervising on a mill blueprint. Yet after the mill is cancelled – note that another mill is built and more on that later – you can see that the IO stays in the supervised state.
I cannot remember whether I redirected the IO to supervise the new mill. If so, then the bug would be that the IO displays supervising state but does not actually confer supervision effect (the clip is not long enough to show that sorry, you need to take my word for it). It’s hard to eliminate this possibility because I am not confident that the replay always record every command. (I have run into cases where it does not.)