The only problem with remaking logic bricks into visual Python components is that a lot of people might potentially get a dramatic decrease in performance in their projects (especially with more intensive tasks such as pathfinding).
Logic bricks are currently written in C++, moving them to Python would perhaps make the BGE the slowest engine available. Part of this is because of the Python lead developer’s personal software ideology that states “it’s fast enough for me”, so there’s little to no effort in making it faster.
The only way this could be done is if this was done in conjunction with finding a way to bring in something like PyPy or Cpython while not requiring people to download an external editor. If not and if the bricks do become Python based, you might as well go ahead and remove the BGE from Blender altogether as even the diehards may not want to use it anymore (in other words, the bricks become much slower and the BGE ends up not being much of an actual ‘game’ engine anymore).
Now one can change the world environment in real-time without futzing around with the video texture module or 2D filter code. Perhaps the BGE will become a competent solution again (or at least good enough to finish existing projects).