To a distant observer, can an object that falls toward a black hole
always return to the starting point?
In the frame of the external observer the infalling observer is always outside the horizon since it takes an infinite amount of coordinate time to fall in. Therefore he could in principle always decide to fly back, given it has an appropriate propulsion system.
In practice it depends on the time it would take the infalling observer to turn on its rocket, since it crosses the horizon in finite proper time, let's say at τ=1, if it hasn't turned on its rocket at, say, τ=0.999, it will not have enough proper time left to turn it on before it is already too late.
In the frame of the outside observer that moment gets stretched infinitely long, but for the infalling object it is a very short period, so if the outside observer sees the infalling object frozen on the horizon without it having it's rocket turned on he will know that the worldline of the object will most probably end in the black hole.
In other words: if the outside observer observes the infalling observer's clock frozen 1 second before τ=1, but knows that it would take 2 seconds of proper time to turn on the engine, he knows that the infalling observer will not make it back.
Isn' t that a good case for hollow black holes?
No, why should it, most infalling matter does not have a rocket attached to it so it freely falls in. In the frame of the outside observer the infalling material asymptotically slows down before it hits the horizon, but the material that was already inside the star before it collapsed is still inside it.