If I have entangled particles A and B, and then B interacts with and becomes entangled with particle C, can that interaction not qualify as an "observation" that collapses the wave function of A + B and de-entangles them?
Or if you prefer to think of balls; if A hits B, and then B hits C, then C's future trajectory is still (forever) affected by what A was doing, as is the trajectory of anything else it hits.