Event horizons without singularities Someone answered this question by saying that black hole entropy conditions and no-hair theorems are asymptotic in nature -- the equations give an ideal solution which is approached quickly but never actually reached from the point of view of an observer outside the event horizon.
Since then I've been wondering whether singularities are ever really created, and if not, why do we worry about naked singularities?
Quick recap: to an external observer, an object falling into a black hole experiences time dilation such that it appears to take an infinite amount of time to cross the event horizon and ends up sitting frozen at the border.
So here's my reasoning: the above should also apply during the formation of the black hole in the first place. The gravitational field approaches infinite density as the constituent matter approaches a central point, but to an outside observer, it takes an infinite amount of time for the singularity to form. In other words, it never happens.
As I understand it, naked singularities are dismissed with hand-waving, "we'll fix it when we go quantum," but I don't see that as necessary. It seems to me that singularities never actually form, although event horizons clearly do.
Does this mean that we can stop worrying? What happens in naked singularity scenarios when there is no singularity yet?
 A: Modelling the formation of a realistic black hole can only be done numerically as the process is far too complicated for an analytic solution to exist. However there is a simplified metric for a collapsing non-rotating star called the Oppenheimer-Snyder metric and this captures the basic principles even though it is too simple to be physically realistic.
In the the Oppenheimer-Snyder metric the event horizon appears first at the centre of the star and then grows outwards until it passes the surface, at which point the collapsing star in entirely inside the horizon. The star then completes its collapse into a singularity, and this happens in finite proper time. So if you were on the surface of the star you'd meet your end at the singularity in finite time as recorded by your wrist watch.
It is true that for an external observer the event horizon never forms because it takes an infinite time as measured by the external observer's clock, and so the singularity never forms either. But to use this to claim the singularity never forms is to treat the external observer's time as somehow specially privileged and this goes against the spirit of GR. We should regard all observers are equal, and since the singularity does form in a finite time for the observer falling inwards with the star, it seems reasonable to claim the singularity does form.
It is true that in a universe of a finite age no observer will ever observe an event horizon, or indeed any singularity naked or otherwise, so in this sense we can "stop worrying". What worries physicists is whether the prediction of naked singularities implies some fundamental problem in general relativity. A naked singularity would imply a breakdown of causality and this it seems worrying that a theory which does such a good job of matching experimental observations could predict something that seems at odds with our expectations, even if we could never do an experiment to observe it.
Whether you could form a naked singularity in a finite coordinate time is an interesting question and I don't know the answer. In principle you could start with a rotating black hole, or more precisely an object that is almost but not quite a black hole, and fling mass in to speed up the rotation until it became extremal. However I don't know whether this could be done in a finite time as measured by the external observer.
A: The currently accepted answer is entirely wrong. It says

Black holes don't start from a point in (for example) the centre of a collapsing star and grow outwards. It's actually the opposite - the event horizon forms outside the collapsing star.

The idea seems to be that the matter that initially collapses to make the black hole doesn't pass through the event horizon and so doesn't "freeze" there.
None of that is true. In an isotropic collapse, the event horizon does start from a point in the center and grow outwards (at the speed of light). In a general collapse, it starts at a one-dimensional set of spacelike separated points called the crease set. In any case, all of the collapsing matter crosses the horizon and in principle appears "frozen" near the horizon forever. (In practice it can't be seen because the redshift is far too large.)
The event horizon is the boundary of the black hole. Like any boundary, it has no gaps in it, because a gap would make the division into inside and outside meaningless. Everything that ends up inside the black hole crosses the boundary.


black hole entropy conditions and no-hair theorems are asymptotic in nature [...] Since then I've been wondering whether singularities are ever really created, and if not, why do we worry about naked singularities?

Technically, everything is asymptotic in nature. If you make waves on the surface of a perfectly still lake, the amplitude of the waves dies down over time but the lake will never be precisely still again. The settling-down time of black holes should be viewed in the same way. This has no bearing on whether the bottom of the lake exists, nor on whether the interior of a black hole exists.
Also, naked singularities are by definition singularities that aren't hidden by an event horizon, so the behavior of the horizon is even less relevant. My impression is that "naked" in the question is just an intensifier (especially given the title), so I'll ignore it.

to an outside observer, it takes an infinite amount of time for the singularity to form. In other words, it never happens.

For a classical black hole (with an event horizon), you can pick a time coordinate that respects causality (anything that happens at $t_1$ can only affect what happens at $t_2$ if $t_1<t_2$), and that covers everything that happens outside of the hole into the indefinite future, and that doesn't cover the black hole interior, or doesn't cover a part of the interior that includes the singularity. From the perspective of that definition of time, there is never a singularity. In effect, the spacetime never gets around to collapsing that last little bit.
But all that you've really done, if you do that, is fail to cover a portion of the spacetime with your coordinate system. That doesn't make it go away. Even if you aren't interested in what happens there, other people are, and the problem isn't solved for them.
A: Yes, the no-hair theorems are asymptotic, and yes a frozen star or red hole approach has some accuracies associated with it.  
However, the issue of cosmic censorship is not well understood, and is still relevant to your situation.
In particular, if a collapsing system forms a singularity that is not naked, then the event horizon could introduce long range asymptotics that make the inside irrelevant to an external observer.  But what if the singularity forms without an event horizon?  Then the singularity is real.

Since then I've been wondering whether singularities are ever really created, and if not, why do we worry about naked singularities?

A naked singularity is exactly the singularity you want to worry about, because it would sit on your past light cone and causally affect you, even though the equations don't make clear predictions about what happens.  So you get a full loss of determinism and causality if you have a naked singularities.
The clothed singularities, as you reference, might not be anything we need to worry at all until we get to a quantum theory or consider other ways to get out of a classical event horizon.

What happens in naked singularity scenarios when there is no singularity yet?

A naked singularity is a singularity that is not inside an event horizon.  So there is no time dilation to save you from seeing it form.  And so if it is naked, then there is no "not yet", it happened, and you can see things coming from the singularity all the way to your eyeball (or the theory itself is just wrong, which either way is pretty bad for the theory).
A: no proof of a singularity can exist as any thing that enters an event horizon can not return as there accessible future only exists in the black hole other wise they have to travel faster than the speed of light which according to Einstein is always the same.
A: The horizons also cannot appear, not only singularities.

Our primary result, that no event horizon forms in gravitational
collapse as seen by an asymptotic observer is suggestive of the
possibility of using the number of local event horizons to classify
and divide Hilbert space into superselection sectors, labeled by the
number of local event horizons. Our result suggests that no operator
could increase the number of event horizons, but the possibility
of reducing the number of pre-existing primordial event horizons is
not so clear and would require that Hawking radiation not cause any
primordial black hole event horizons to evaporate completely.

source
See also this answer:
https://physics.stackexchange.com/a/21357/1186
