This is an oft used technique in gamedev, where instead of print statements we will print out visualisations to the screen, such as bounding boxes, trajectory lines, etc.
He describes that the visualizations are generated by logging the relevant events from a running program, so I'm not sure what you mean by extending it to render based on the "actual runtime of a real program".
I have a theory that much complexity could be understood if humans could just see the interactions and the data flowing.
Even if the program took 100X longer to run, if you can see the bug, you can fix it. If you can see the complexity, you can understand it.