Here is my understanding of why this optimization is faulty:

1. we traverse along the grid
2. the dark blue box is where we first see something "interesting" -- i.e. a circle
3. we calculate the intersection based on what we saw in the blue box and the circle
4. the intersection we calculated, the red dot, is NOT the nearest
5. instead, the intersection calculation should be done in the next grid we'd move to -- the fix for this is as on the slides, checking if the intersection point is in our current cell

I'm still a little confused about this edge case. Does our algorithm look for collisions of bounding boxes within the grid first, then just looks at that box of interest afterwards? Why isn't there a grid collision with that nearer, closer bounding volume? Wouldn't that stop us from incorrectly marking that red part as the nearest?