Relativistic charge density in a closed loop When charges of conductance are at rest, there is an average distance between them. The relativistic origin of magnetic field says that distances between electrons shrink when they are set into a motion (we get current). This means that the electron density is increased while total charge is conserved. I just try to relate the contracted distance between electrons with the conductor, which stands still and does not contract.
If we had infinite-long wire, we could borrow any amount of charge from the edges of Hotel Infinity to increase the electron density without increasing it. But, real wires are of limited length. So, condensing charge -*--*--*--*--*--*- => -----*-*-*-*-*--- cannot make one segment more dense without depleting density in the other parts. I cannot make the whole wire more charged (and keep it neutral in the lab frame, at that). The same applies to the loop. 
 * - * - * - *      
 |           |      * * * *
 *           * ==>  *     *
 |           |      * * * *
 * - * - * - *       

Here alternative physics draw a nicer picture to ask the same question

The image says that the loop of electrons shrinks as they start to flow. Yet, mechanically, electrons do not escape the solid wires, which stand still and do not contract! 
That is the question: how do inter-electron distances shrink without leading to logical absurd?
This question stems from understanding why loop with current stays neutral, despite the electron density increase. I first guessed that the conductor stays neutral because positive charges flow symmetrically in opposite direction so that relativistic charge density increase in both directions compensate each other and keep the loop neutral. In this case the conducting loop would contract proportionally with the loop of electrons but it would not stay at rest in this case. Yet, the problem is that the loop of wire stays at rest and does not contract.
 A: I was told in the paradox of conductor's neutrality and this is confirmed in Follow-Up #2: relativity and charge density that in addition to contraction, mentioned in the relativistic electromagnetism tutorials, the electrons of conduction experience an expanding force, when accelerated. That is they see that the distances between them stretch out in their proper frame, as they experience the acceleration. The expansion is natural for non-rigid bodies, as explained in the Bell's spaceship paradox. 
I emphasize this since it is absolutely counter-intuitive for any layman like me, who are taught that moving train is contracted w.r.t. stationary observer. It is underspoken rigidness of the train that prevents it from expansion in its proper frame, so the train shrinks indeed in the stationary frame when accelerates. But, spacing between electrons of conduction is not tied rigidly, so, when electrons experience acceleration, they also see that the distances between them increase. The expansion factor is exactly Lorentz $\gamma$, which exactly compensates the contraction seen from the stationary station. The net effect is that distance between electrons is not changed in the lab frame. So, maintaining neutrality also eliminates the paradox of wire neutrality. 
Though I still do not understand the source of the expanding force (which energy does support it?), the paradox vanished. Might be this also partially answers the Relativistic origin of magnetic field. 
However, one thing is still not clear. If electrons see their chain expanded while loop contracted then how do they fit into the loop? I guess that it's a famous ladder paradox. Yet, it is in a loop now and since my question is exactly about this case, I would like to resolve it also.
A: Distances are shrunk, but not in all directions equally!
One should read about addition of relativistic velocities for non-parallel directions.
In the case of a linear wire, in which the electrons move in the same direction (and even same speed) as the test charge things are simplified a lot so that equations are relative easy.
However for non-parallel (or non equal) velocities things are just that bit different.
Your loop will look somewhat like this (depending on the orientation and position of the loop with regard to the velocity vector of the testcharge):
--+-+-++
-      +
-      +
-      +
-      +
--+-+-++

A: After reading several textbook examples relating to this matter I see how you question arises. Most textbooks simply start with a current in a (squared) ring, and assert that the ring is neutral in the lab frame. Then they view the ring from the moving frame of the testcharge.
None of the textbooks care to explain the internal workings of this trivial neutrally charged, current carrying ring. Hence your question, I presume...
By the way, someone else came with a similar question, but also did not get a fully satisfying answer. http://van.physics.illinois.edu/qa/listing.php?id=2358
You are correct to assume that the solution you suggest, ie. only moving electrons, and a neutral wire, will lead to a logical absurd.
Let me therefore suggest another option that you might embrace: What if a current is not merely electrons moving one way, but also positively charged holes the other way at the same velocity? Both would contract in the same way, and the ring could therefore stay neutral (in the lab frame only).
You may read about the concept of holes: (normally only used for valance band conduction)
http://www.allaboutcircuits.com/vol_3/chpt_2/5.html
Another solution might be that when a current (in this case only moving electrons) starts, it will not start instantaneously (the electromagnetic field needs time to propagate) along the whole wire : The electrons that are closest to the positive pole of the battery will start moving first, and thus at first their density will be a bit lower. Like cars after a traffic congestion, their distances increase. Maybe they increase just as much as the relativistic contraction does contract, keeping the total density the same. (I've not taken the effort to do this in formula's)
Let me know what you think about it.
A: The problem of charge conservation is much simpler than stated, if you consider that current always needs a closed circuit and that the electron gas is compressible. Firstly you will note that all charge disparities in a current carrying wire observed by a moving observer belong to an electric dipole field like that of a rectangular circuit where its upper section might charge up positive and its opposed lower section negative summing up to Zero total charge difference.
As to the Lorentz-contraction of an electron gas drifting truogh a current carrying ring it is obvious that the solid state structure of the ring is capable to inhibit that Lorentz contraction mechanically, which however results in a very small mechanical compressive stress appearing within such ring. If you look at the equations of state of an electron gas you can calculate the amont of stress needed to inhibit Lorenz-contraction. Of course the Lorentz-contraction of any elastic body can be inhibited by a counter-acting mechanical stress as given by Hookes Law.  
A: We have to be careful using our classical intuition when the problem isn't classical in nature. The electrons in the conductor are in delocalized states, that are distributed over the conductor. In this picture, it makes no sense to speak about the distance between the electrons. To find the charge in a given region, we should instead sum the probability distribution of each electron for this region. In the rest frame of the conductor the electrons are at rest in average. They have an angular momentum, but their center of mass doesn't move. 
In the wire electrons are circulation in both directions, even when there is no current. A current arises when more electrons are in states circulation in one direction than the other. The situation is much as an electron in an atom. I think this is a better picture to hold in mind.
The question then is, how these electrons states transform when viewed in a different frame?
Unfortunately, I have not yet found a proper analysis of this.
But we know for sure that an electrical force will occur in the rest frame of a charge, that moves through a magnetic field. I believe, however, it is mistaken to think this force arises due to length contraction alone. It is an electrodynamic effect, and not an electrostatic. See my questions about this:
Is magnetism an electrostatic or electrodynamic effect in the rest frame of the affected charge?
How do you model the electric field of an electron in motion in a conductor? Is the field given by static or dynamic equations?
A: I think your question is very relevant.
I think I can put it in a wider perspective.
In Minkowski spacetime when you close a loop interesting things happen.
I believe the ring configuration that you point out exposes a severe weakness in the usual exposition of magnetism-as-relativistic-side-effect-of-the-coulomb-force.
I will abbreviate this 'Relativistic Charge Density Contraction' as 'RCDC'.
The RCDC explanation does not pay attention to relativity of simultaneity. That is a perilous simplification. It may be that the RCDC explanation contains errors that fortuitously drop away against each other. While the RCDC explanation produces the sought after result, it may nevertheless be unsound. 
About loop geometry in special relativity:
I find the article Sagnac effect, twin paradox and Ehrenfest paradox by the physicist Olaf Wucknitz very interesting. Wucknitz explores what you get when you close a loop in Minkowski spacetime. It may help you understand why the ring configuration that you point out poses a challenge to the RCDC explanation. 
