How do quantum fluctuations become gravity wells? Restatement of a question that was never answered completely:
How do quantum fluctuations become gravity wells? I thought the whole idea of the fluctuation was that it had to happen so quickly that the universe didn't notice. I see how a momentary change in a field can stick around long enough to do actual work, like creating a potential for a gravity well. 
 A: Quantum fluctuations are turned into large-scale cosmological perturbations in the energy density by cosmic inflation.  These large scale energy density perturbations can be interpreted as gravity wells, and the perturbations will lead to further inhomogeneities in the energy density as these gravity wells attract more surrounding energy together into clumps.
Ordinarily, yes, quantum vacuum fluctuations happen on very short time scales.  The longer the time scale the smaller the energy fluctuations and the closer any measured energy density has to be to the expected value of the vacuum energy.
However, during the inflationary phase of the early universe, space is expanding rapidly and exponentially (rather than approximately linearly, with only slight acceleration, as it is currently).  So slight fluctuations get amplified very quickly.  There is a Hubble horizon (H) during inflation (also called a deSitter horizon, in this context).  Any distances further apart than this horizon are not a part of the same "observable universe" because space is expanding faster than light can make it to the horizon.  This is also very similar to the event horizon of a black hole.  In comoving coordinates, this event horizon is actually shrinking.  Less and less of the universe is in causal contact during the inflationary phase (and then afterwards, more and more of it is recovered as the horizon expands again).
Now consider quantum fluctuations that happen near this event horizon.  They are similar to the vacuum fluctuations that happen near the event horizon of a black hole.  If a particle and an antiparticle are created right on the horizon, they may end up on opposite sides.  So what started out as a virtual quantum fluctuation ends up looking like a real thermal fluctuation.  In this way, inflation converts quantum fluctuations into something that looks like ordinary cosmological fluctuations after everything is said and done.
So yes, ordinarily they have to happen faster than the universe can notice.  But if the universe is expanding even faster (and the accessible part of the universe in comoving coordinates is shrinking fast enough), it has no choice but to notice.  A temporary fluctuation has been amplified and has become permanent.
A: Doonesbury: I believe your question only refers to the "gravity wells" which have been proposed to explain the observed "clumping" of matter and radiation energy in the observed universe. The distribution of these "wells" (more like wrinkles, or valleys) appears to follow a quantum kind of pattern. "Normally," no quantum effect is "large" enough to affect macrostructures, but there is a theory which accounts for this very large scale "quantum" distribution. This theory utilizes a special scalar field called the "inflaton" which accounts for the translation of potential energy represented "continuously" (linearly) across the early "universe" into mass particles and radiation energy that was distributed in the apparently nonlinear (non-uniform) distribution of matter (clumping) in the observed universe. (So, it explains the DISTRIBUTION of the matter and energy, NOT the "amount.") This distribution appears to be in a quantum style pattern.
These areas of concentration of matter CAN also be described (how about "mapped?" does that help?) in the same way as the smaller gravity "wells" generated by massive objects like stars and galaxies. The PATTERN of distribution APPEARS to have been imposed by a quantum interaction translated via the Inflaton field. How, you ask, if such quantum effects are normally too "small?"
According to Inflaton Theory, the early universe may have been perturbed by a quantum fluctuation at a veeeerrrrrry early "moment" when the scale of interaction just happened to be such that the quantum nature of the perturbation was comparatively "large" enough to cause this translation of potential energy to occur over the whole universe in a non-linear (or nonuniform, depending on your mathematical point of view) fashion. Such a quantum distribution of matter effect, according to the relevant calculations, could only have taken place at such an early moment, and could not, for example, happen "now." That's because, at the early moment, "space" was curved very tightly, and the "universe" was very tiny (at more or less the quantum scale)—as opposed to now, when it has become nearly "flat" and ginormously "big."
Think about scalar fields this way: suppose you are about to shoot yourself from a cannon. A tiny push on one side of the cannon can make a huge difference in your ultimate destination—IF the push is done BEFORE you fire the cannon, or even, a split second after. That is, when your total trajectory is comparably tiny. But, the longer you travel on your trajectory (the more your path has "inflated") the less total effect such a push (which can be described mathematically as a "field") will have on your destination. The scale of the push has become dwarfed by the scale of the trajectory you have traveled. In a non-scalar relationship, the push would have inflated in the same proportion as your trajectory so as to result in the same total deflection. (I apologize for the trigonometric inaccuracy here; just think of the total deflection being the same no matter how long the trajectory.) Quantum fluctuations are, so far as we can tell, QUITE scalar, and therefore, such a theory DEPENDS on the universe having been such and such a "radius" with such and such a curvature in order for the perturbation to have been so "large" in its effect. 
Or, to put it another way, the Inflaton Field is the "how" and the timing determines the "how much." 'Twas a one-shot deal. If it had not "happened," then, perhaps, matter and energy would would have been distributed more or less uniformly. Pretty, but boring.
The Inflaton theory does adequately account for the observed clumping, even if it does toss yet one more field into the soup. In other words, it COULD have happened that way. Not everyone agrees that it DID happen that way, or that it is the "only" way it could have happened. 
So far as the question of "conservation" goes, it's not that some matter and energy was "removed" from the universe according to a quantum shaped "pattern;" it was merely "clumped" according to a quantum-shaped "pattern." Does that help? (If on the other hand, you are asking, "where is all the energy lost due to redshift going?" then it MAY be translated into the Gravity energy potentials as recently observed in gravity "waves." Maybe.) 
And, yeah, Hawking also showed that quantum scale gravitational effects can occur veeerrrrrrrrryyyy near event horizons, too. But I don't think I'd chracterize those as gravity wells; I'd say, they happen NEAR gravity wells. But I could be wrong.
A: A gravity well is a mathematical description of gravity in Newtonian physics. General Relativity equations reduce to Newtonian equations in flat space, i.e. the usual 1/r gravitational potential appears when space has expanded enough that it could be considered flat. 
In the Big Bang model at present, in order to explain the homogeneity seen in the Cosmic Microwave Background radiation the inflationary model has been proposed , as described by the other answers.
At a time before 10^-32 seconds from the singularity of the General Relativity solution, which was the original Big Bang model that modeled the Hubble expansion, a quantum regime is assumed with a specific field ,the  inflaton, not seen in our laboratory experiments and not predicted by the particle physics Standard Model. It is indeed a form of the Heisenberg Uncertainty that appears in the behavior of this field.
I call it a form because the commutators that describe the mathematics of the HUP in flat space will have a different mathematical formulation due to General Relativity, but the concept of uncertainty is the same, and is what powers the homogeneity during that period: time and space are probabilities and not consecutive manifestations for the inflaton and thus the light cone constraints, which are needed in a thermodynamic model of the universe,  are nonexistent at the inflation times.

The Big Bang model as it is at present. The whole period before 10^-32 is presumed to be a homogenizing period with very high General Relativity controlled curvatures. Gravity wells have no meaning there, but density fluctuations of the stress energy GR tensor have a meaning. As the expansion continues these density fluctuations develop  in the CMB map.
By the time the quantum age ends and protons have formed, local spaces are very close to flat spaces, and General Relativity locally is identical with the Newtonian physics framework. ( see here for how this happens)This means there are more protons and electrons concentrated where the density fluctuations were high and less in between. The Newtonian gravity wells will appear naturally , because of the 1/r attraction of the gravitational potential, clusters will be pulled to their center of mass forming galaxies and clusters of galaxies as we observe them.
