That isn't exactly accurate. There's no "modes" of Panda3D, its simply how the Panda3D engine is interfaced with to render the game. Panda3D is a "library" of many different "functions" or commands you could say. Being that Panda3D's purpose is for creating games it has a lot of game related functions for rendering a scene, lighting, shadows, etc. These functions are all called and written into code that the game developer themselves write, at the basic level it is just done with inserting objects into a scene and having Panda3D handle the lighting, shadows, etc. But on more advanced side of things developers can write custom shaders and lighting, implement shadows, bump mapping, and everything else Panda3D has functionality for.
TLOPO already has their code that interfaces with the engine written (most of it being from POTCO), as well as custom shaders and such. It would be much more involved to implement a better system than what they have. Likely a quite extensive partial rewrite of some systems. Not to mention Panda3D really struggles to utilize GPUs (Graphics cards) to their full extent, so either they would have to use some trickery to get that to work, to rewrite some of Panda3D itself, or just let the CPU do the majority of the work as it is now, and that's not a great idea.
As for what you saw, in those videos. Some of them were custom or heavily modified Panda3D rendering code, not very many objects on screen to render, or simply ultra optimized to that specific scene. It's relatively easy to control for performance in a unchanging scene. But in the real world applications you must take into account many different variables and it gets messy.
