In fact, the entire interior of the event horizon is actually physically invalid in these simulations. The formalisms used trap the errors inside the event horizon, as the errors turn out to be strictly causal. And because of that, theoretically they can't escape

Of course that analysis breaks down in the face of discretisation, so errors tend to leak out a bit under low resolutions, so you have to handle things pretty careful. Either way, you shouldn't draw any conclusions about the interior

Source: I've done a lot of these simulations