I'm going to assume some things that are pretty well accepted by the physics community. Of course, one can always find people with opposing viewpoints.
I attended a talk on the firewall issue by Leonard Susskind last week, and he started with some interesting comments on the whole "what do theoretical physicists do?" question.
He gave four cases:
The conflict with firewalls is that quantum entanglement (which has held up very well so far) shouldn't cause the equivalence principal to be violated (this, too, has done very well experimentally). The equivalence principal states that an accelerated observer, absent other information, can't tell if their in a rocket or standing on a surface in a gravitational field. Implied by this is the "no drama" notion that says that nothing interesting should happen when one falls through an event horizon, which itself is a smooth bit of space-time. (I'm assuming here, for the sake of a macroscopic observer, that it's a big enough black hole that tides don't come into play until well towards the central singularity and that the surroundings aren't full of super heated, radiating matter.) The firewall hypothesis arises as a possible solution to what happens (very) late in the evolution of a black hole when most of the matter still inside the horizon is entangled with matter that's been emitted as Hawking radiation. The equivalence principal says that a firewall, being very dramatic, shouldn't happen. This firewall isn't the same as the very, very late stage of a black hole when the Hawking radiation is so intense that nothing is likely to get past and make its way into the hole. Maldacena and Susskind seek to resolve this and have come up with the notion that EPR bridges (entanglement) and wormholes (general relativity) are the same thing. (Now before everyone gets going about wormholes, these aren't expected to be anything more than a sort of identity mapping between entangled particles.) I don't claim to follow everything about how the initial entanglement described in the paper actually comes about, but the overall argument has a feeling of making sense, and a room full of gray haired physicists didn't tear it down. Susskind also pointed out that if black hole horizons become messy, so do other kinds of horizon such as cosmological ones, adding further inelegant complications for the theorists.
The paper by Ellis is interesting in that it could just make the whole problem go away along with the information paradox. Ellis's argument, however, assumes that the classical event horizon that goes into the infinite future is the real one. Personally, I like dynamical/apparent horizons better.
The value of a program is proportional to the weight of its output.