It is useful for making other nodes follow a path, without coding the movement pattern. For that, the nodes must be children of this node. The descendant nodes will then move accordingly when setting the [member progress] in this node.
The points along the [Curve2D] of the [Path2D] are precomputed before use, for faster calculations. The point at the requested offset is then calculated interpolating between two adjacent cached points. This may present a problem if the curve makes sharp turns, as the cached points may not follow the curve closely enough.
There are two answers to this problem: either increase the number of cached points and increase memory consumption, or make a cubic interpolation between two points at the cost of (slightly) slower calculations.
The distance along the path as a number in the range 0.0 (for the first vertex) to 1.0 (for the last). This is just another way of expressing the progress within the path, as the offset supplied is multiplied internally by the path's length.