Would it help to use multiple sensors at different positions and use the common points to filter out the 2nd+ bounces?

I think this would work for, say, a mirror but what about something like brushed metal?

Are those other bounces scattered enough that multiple perspective still produce an error?