What if Mary is Chinese? :P

I mean, what if there is a person not understanding Chinese in a room, operating the Chinese symbols they don't understand, according to formal rules that make no sense to them. The system already "knows" (on the symbol level, not the person who operates it) everything about the "red" color, but it has never perceived the red color in its input. And then, one day, it receives the red color in the input. If there is an unusual response by the system, what exactly caused it? (For extra layer of complication, let's suppose that the inputs to the "Chinese room" are bit streams containing JPEG images, so even the person operating the room has never seen the red color.)

To add more context, what if the perceived red object is a trolley running down the railway track...