Time symmetry and general relativity

Classical general relativity is time-symmetric: flip the $t$ direction on your space-time manifold, and you get another valid solution to the field equations.

The difference is that in this picture the black hole is replaced with a white hole occasionally spewing our particles... or metal balls. Nothing can get into it.

The weird arbitrariness of white holes spewing out non-random things hints that something is not right. In normal life things are not time-symmetric, and we typically attribute this to thermodynamics: entropy typically increases and we cannot undo past events without spending considerable (entropy increasing) resources.

This similarity between black holes always absorbing things and entropy always increasing led Hawking and others to suggest black hole thermodynamics: black holes behave like thermodynamic systems with temperature, and this is why the time-reversed case does not happen. But black hole thermodynamics is not classical physics: extra assumptions have been introduced, and many think that quantum gravity will make it clearer what is going on.

I am far from an expert but I think I can help.

The arrow of time is a macroscopic phenomenon built on top of completely symmetric dynamical laws of physics held through CPT-symmetry. Where does the arrow of time come from then, why is there observed T-assymetry? First note CPT together is held, witnessing T-assymetry alone is not a problem, and we see it all the time, usually as the second law. T-symmetry is only violated in the dynamical laws (in the dynamical laws themselves, not talking about the macroscopic probability of second law which is another t-asymmetry from the laws themselves) in two narrow cases, "one through the mixing of different flavours of quarks in their weak decays, the second through a direct CP violation in strong interactions. The first is seen in experiments, the second is strongly constrained by the non-observation of the EDM of a neutron." [1] But neither of these two dynamics is important to your question, and this dynamical T-assymetry is NOT why we have the macroscopic second law anyway. In fact, we have a second law because "the constant increase of entropy we observe happens only because of the initial state of our universe." [1] The low entropy big bang prior condition is taken as a given in modern physics, and the second law is a consequence of that fact plus the dynamical laws we know and love acting upon those conditions.

Onto GR specifically. "But because the equations of general relativity are time-reversible – they exhibit Time reversal symmetry – general relativity must also allow the time-reverse of this type of "realistic" black hole that forms from collapsing matter. The time-reversed case would be a white hole that has existed since the beginning of the universe, and which emits matter until it finally "explodes" and disappears.[10] Despite the fact that such objects are permitted theoretically, they are not taken as seriously as black holes by physicists, since there would be no processes that would naturally lead to their formation; they could exist only if they were built into the initial conditions of the Big Bang." [2]

So just because a white whole is a T reversible solution in the equations of GR does not mean it has to or can occur. In fact we think they don't occur because they'd have to be around at the big bang, and be stable since. We can't and shouldn't expect to see every naively allowable solution to the equations of physics without taking initial conditions/cosmology into account. And we assume a hot big bang prior condition.

Lastly, about playing a physics video in reverse. Your setup would still betray an arrow of time. The smaller ball will have greater velocity and acceleration in one direction due to the planet's own gravity. And knowing what I know about physics, I bet I can pick out which direction is the "natural" one. I would see the ball accelerating away from the planet in one version and instantly know due to the second law and fluctuation theorem (exponentially decreasing odds to witness reversing of entropy further a system is from non-equilibrium). Thus I am almost positively seeing the reversed video. The only way this works is for an extremely small setup, where the second law (which is describes the macroscopic average) doesn't have a large role. Like a single photon in an idealized box.

So in summary, you can't expect the exact time reversed macroscopic behavior in a world with the second law, even outside of the dynamical t-assymetry in the strong and weak force.