Seeking intuitive explanations regarding $A_{n-1}$ singularities I am studying Ben Craps' lecture notes Big Bang Models In String Theory for my undergraduate thesis. I am not well versed in topology, and would like intuitive explanations for certain ideas regarding Asymptotically Locally Euclidean (ALE) singularities.
Firstly, the author states.

The orbifold has a singularity at the ﬁxed point $(z_{1}, z_{2}) = (0, 0)$; mathematically, it is known as an $A_{n−1}$ singularity, a special case of an ALE singularity. Once again, perturbative string theory turns out to be completely smooth due to the twisted closed strings. It is well-known that geometrically, the $A_{n−1}$ singularity can be resolved into $n−1$ intersecting two-spheres. Before saying more about this singularity, I have to brieﬂy introduce the concept of D-branes.

What does it mean for a singularity to be resolved into $n-1$ intersecting two spheres?
Then the author introduces two types of D-branes: bulk and fractional.

For the orbifolds we are considering, one ﬁnds that there exist two types of
D-branes. The ﬁrst type are “bulk” D-branes. On the covering space, they correspond to
a $\mathbb{Z}_n$ symmetric conﬁguration of $n$ D-branes. Bulk branes have the property that they can move anywhere in the orbifold: the images move in such a way that the conﬁguration remains symmetric under the $\mathbb{Z}_n$ orbifold group.

What is the covering space of an orbifold, intuitively? Are these images under the orbifold identification of the brane?

The second type are “fractional” D-branes.
On the covering space, they correspond to just a single D-brane placed at the ﬁxed point,
which is a symmetric conﬁguration by itself. Fractional branes are stuck at the orbifold sin-
gularity: to move away from it, they would need $n−1$ companions to keep the conﬁguration
symmetric, but those companions aren’t there. The fact that fractional branes are stuck at
the singularity makes them ideal probes of the structure of the singularity.We have seen
before that an $A_{n−1}$ singularity can be viewed as a limit of the resolved $A_{n−1}$ singularity, where the $n − 1$ two-spheres collapse into a single point.

I don't understand what he means by needing $n-1$ companions to keep a configuration symmetric.
 A: What does it mean for a singularity to be resolved into n−1 intersecting two spheres?
Just to be sure, the mathematically precise of idea of "resolving a singularity" is given in the realm of algebraic geometry under the name blow up. Now, what is a blow up intuitively speaking? Algebraic geometers have sets of rules to replace singular spaces transmuting them into smooth manifolds in (a more or less) canonical way.
Let's analize the $TN_{k}$ ($k$-centered Taub-NUT space) case. Its singularities are arranged in the pattern of the roots of the $A_{k-1}$ Lie algebra. You must think on those singularities as points where the curvature becomes infinite. Resolving them just means replacing points by spheres of finite size (smaller that the separation between points) making the background smooth.
Resolving the base $\Rightarrow$
Is this replacement cannonical in some way? It is. The general rule is to replace simple classes of singularities by the projectivization of its normal bundle. In our case the projectivization of a point in $\mathbb{R}^{3}$ is a 2-sphere.
I urge you to read the classic Strings on orbifolds to understand how string theory is consistent over backgrounds with orbifold singularities. The classic McKay correspondence to learn about blow ups, Orbifold Resolution by D-Branes
 to understand in full detail the physics of the stringy resolution of orbifold singularities and my answer for a roadmap to interesting applications of orbifolds in string theory.
What is the covering space of an orbifold, intuitively? Are these images under the orbifold identification of the brane? You can think on the covering space of an orbifold is any of its smooth resolutions.
I don't understand what he means by needing n−1 companions to keep a configuration symmetric. Imagine that you resolve just one point-singularity among a large number of them, leaving the rest of them untouched. That's not a valid operation in the theory of orbifolds. The reason is that any allowed birrational transformation must preserve the action of $\mathbb{Z}_{n}$ intact. Something that the sort I mentioned does not treat democratically all the symmetry generators (a symmetry breaking is taking place) and that's not what we want, the paper you are reading is interested in exploring the stringy dynamics of a particular spacetime without spoiling its symmetries.
In plain english: Any transformation made over a point singularity must be performed in exactly the same way for over any other one. Otherwise the global symmetry group $\mathbb{Z}_{n}$ is broken.
