The Snap action provides the classic way of window snapping which is a de-facto standard in Unix/Linux window managers; another name of this feature is "magnetic/sticky borders".
It works when you drag or size a window with the mouse: when any of window's borders comes to such visual bounds as desktop/monitor boundaries or other windows closer than the specified amount of pixels - the border automatically "sticks" (or "snaps") for some time to that bound. If you continue to drag/size - the stuck window's border will "unstick" from the bound it was stuck to and then will go freely until you drag it to another bound.
Further advancement of this action is dragging/sizing a window not continuously but gradually with some defined step (e.g. 10 pixels, 20 pixels, etc.) - as if you'd move a window along the virtual grid, and the window would snap to that grid's cells.
Hint This action can greatly simplify and ease the momentary allocation of windows - it frees you from the over-scrupulous mouse manipulations when you try to place some windows side-by-side without both overlapping and gaps or place a window exactly at the desktop boundary. In most cases, this is exactly what you want to get so that classic snapping makes window dragging/sizing far more intuitive and convenient.
Hint While dragging a window, you can temporarily disable this kind of snapping by pressing and holding the defined key combination; release it to activate the snapping back.