[Please tell me if the following is incomprehensible, boring, erroneous, interesting or intriguing (etc.) ]
For a long while I've been aware of an argument that suggests though the mechanism for Hawking radiation makes perfect sense, it will not ever be observed by someone at a distance from a black hole, but I think I've finally convinced myself it will be (unsurprisingly - Hawking is better than me at this stuff ). But in doing so, it makes it clear just how weird the extremes of general relativity combined with quantum mechanics can be.
Imagine the simplest case for formation of a black hole, where there is a perfectly symmetrical spherical shell of matter collapsing under its own gravitational attraction. If big enough it would be reasonable to ignore the things that would confuse the perfect collapse (interactions within the material). In fact if really huge the sphere of matter could be quite diffuse and still form a black hole. [Using a spherical shell of matter rather than a more real-worldly solid sphere of matter just makes the discussion a bit simpler]
Anyhow, as far as I know, general relativity predicts that if you watch this from a distance, a strange thing happens. The collapse accelerates as the sphere gets smaller for quite a while, but in addition the sphere gets gravitationally red-shifted, which has the effect of appearing to slow the collapse. At some point to the remote observer the collapse seems to start to slow down, and eventually gets to a state where the shell appears to be almost frozen, massively red-shifted, just above the radius where the event horizon for this much mass would be expected to be. Shrinking the last little bit to the event horizon seems to take literally eternity from the point of view of the remote observer. A key point is that for eternity, every atom in the collapsing matter is (in principle) observable by the observer at a distance.
Now if Hawking radiation is observed at some time in the future, the following process must occur (many, many times). A pair of photons must appear in the space near where the event horizon would be expected to form. The one of the photons with negative energy manages to get past the event horizon in the time given to it by Heisenberg's uncertainty principle (delta t * delta E >= h) to fall into the black hole and the one with positive energy manages to escape from the gravitational field completely. The photon that escaped goes all the way to the remote observer.
But here's a conceptual problem. The remote observer can see all the original matter hovering at a radius just above where the event horizon will form, and now sees this radiation being emitted by the black hole. If the remote observer placed an entire sphere of photon detectors at a large distance from the black hole, according to Hawking's theory, he can expect to see the entire mass of the black hole emitted as black hole radiation over a finite (albeit very long) time. But meanwhile the entire shell of matter is still visible contracting above the event horizon. So how can the effect of the formation of the event horizon get to him even though the event horizon appears never to form, by a straight shortest line of sight view? After the entire mass of the black hole has been transformed to black body radiation, wouldn't it be possible (in principle) to fly a spaceship on a trajectory that takes it very, very close to the event horizon, and to scoop up a little of the mass that formed the black hole. But isn't this paradoxical?
Perhaps the simplest explanation is that the virtual photon pair can form just outside the shell of matter, when it is very, very close to the event horizon. This means the escaping photon starts from a point nearer to the remote observer than the shell of matter, and due to gravitational time dilation, being a little nearer can mean the signal gets to the remote observer a very great time earlier (the shell appears takes an infinite length of time to fall the last micron to the event horizon).
And that's why I now do believe Hawking radiation can be observed.