This would just be easier with a true PhysX fluid simulation correct?

Currently unsupported in 3dgs. Sad.

Well I like txesmi's way.
I was going to sum normals into 3d shapes then apply poly to poly c_moves, get the direction and angles of shape to boat hit-contacts and apply them as force on the boat... Insane !!!

The problem of collision traces is that you need CPU animated vertex for the water model...