That's my object as seen in the editor. It's a ball, RigidBody2D but set to Kinematic mode:
Ball (Collision layer/mask: 3/3)
--Sprite
--CollisionShape2D
--Area2D (CL/M: 1/1)
----CollisionShape2D
Area2D (CL/M: 3/3)
--Sprite
--CollisionShape2D
All CollisionShape2Ds are the same, one on top of the other
There is also KinematicBody2D (CL/M: 1/1) being a cursor (it's like that because there is no standard controller, mouse or keyboard - just my library feeding Vector2s every frame).
What I want is to set Ball's mode to Rigid after I detect cursor inside Ball's Area2D (with some additional requirements) so it falls down and then the goal is to bring it back to the loose Area2D. All the walls are also 3/3. I can't seem to detect the collision. Can I have a child with different collision layer/mask than a parent? Can CollisionShape2D of different layer interfere with detection? Does z-index matter? I'm lost.