E.g. Imagine a player jumps and bumps into a light source in a simple game: We add an up force to the player object, when it bumps into the light source we calculate the collision of the two shapes and calculate the resultant forces, move the player and light, and then recalculate the light of the scene for that frame.

In a complex game: We designed the inverse kinematics system to position the feet without the idea that a player would ever leave the ground, we don't have any more mocap studio time so we don't have a jump animation, the characters hair physics look really weird if we just apply an upwards force so we need to code a work around for that, if the player bumps into the light we can't move it because we have to precalculate the lighting at compile time, etc. Better to just not let the character jump.