It is just a combined movement or are there other advantages?
There’s a type of animation called “cutout animation”.
You can’t do that in a reasonable way without having some sort of child sprites.
You can also use child sprites for spritesheet animation (vs. skeletal “cutout animation”). In both, for example, you can layer a character’s armor sprite on top of a base body sprite. When the character acquires armor, you can show the armor sprite. If the character upgrades armor, you can swap in a different armor spritesheet. In the case of spritesheet animation, the children are typically all synchronized to play the same animations.
But the biggest advantage is that you can separate the object’s behavior from its appearance. The root object can hold all of the behavior components, and child objects can hold sprites. In addition to having separation of concerns, this way you can scale child objects (e.g., give a character a bigger head) or even swap them out (e.g., human to werewolf) by modifying child objects.
That reminds me.
Disgaea games used child sprites for weapons.
The sprites are traditionally animated and can be palette swapped.
However, they hold weapons, and the weapons correctly reflect currently equipped items (at least for humanoid units). This is heavily used in animations, because Disgaea has over the top special attacks that are mini-cutscenes with acrobatics.
That’s one other use.
Then there’s diablo-like games which could display some of the character equipment despite characters being 2d, because the sprites were assembled from pieces.