I think everything depends on the rate at which data is coming in and if you want your plot to reflect that rate. I suspect that even with a few updates per second the matplotlib approach could work so I would definitely try it to see if it meets your needs and just be pragmatic about it 
The only other alternative I can think of is implementing your own plotting functions within PsychoPy, but this is not so trivial. So although this will probably allow you to render a real time graph much faster, it will take a lot of work to implement it (and you also need to implement it in an efficient way in order for it to be faster).