The whole physics simulation system seems to be inconsistent,
here are a few of the things I have come across…
Try adding a child actor to an actor, its physics and collisions are either none existent or totally haywire.
eg a child actor with collision could be pushed by a pawn in one direction but the other direction locked solid… …Yet the actor could move freely through the pawn in both directions and not collide .
Add two components to an actor without physics enabled.
Give each a custom set of collision properties eg hook and chain… select either of them and the correct custom/hook–chain collision properties show
Select both at the same time and you will see that as a pair the so called custom properties are probably listed as world dynamic or world static … and changing these properties as a pair does affect the relative collisions.
Go back to selecting just a single item and the correct ??? collisions are shown.
An actor with a physics material will behave differently to the same actor with a “material” that has a physical material component ( same physical material)
Constraints on a child actor do not constrain (or very weak)
Setting custom collisions and giving it a custom group eg hook or chain ( set up in project settings ) does not give the same result as say …physic body with identical settings
(corresponding collisions from the other way were matched to whichever group was allocated )
Why does giving a component mass (physics disabled) alter its behaviour ?
If physics would work in a consistent and relative manner there would be no limit to the mechanical simulations possible
Paul G
So far I have been able to ( by trial and error) overcome most limitations to produce a purely physics based rail system but methods are extremely inconsistent.