What controls the size of crystals' pyramid-like top? so we are growing crystals from supersaturated salty solutions with the children and it is great fun.
I do have some basic understanding about nucleation and growth, but what is I cannot figure out all is the reason for which say hexagonal prismatic crystals often end with an approximately pyramid-like portion on the top (that is, for a crystal growing upwards, or alternatively, the portion formed last).

I made quite a search and can get plenty of information on "forms", "habits", crystallographic planes, angles, but not on the "why".
Where is the convenience of "closing up" a crystal in such fashion? And what does trigger it? Why would not grow further and postpone the "pyramidisation"?
I appreciate these might be obvious questions, I would be most grateful even only for a reference to look in.
Thanks
EDIT It was clarified in the comments how some assumptions of mine, such as that the pyramid would grow at the end, etc, were plainly wrong. Questions remain though, how does macroscopic crystal growth works? How is one crystallite developing in a pyramid / prism, and then growing? What triggers the change from pyramidal / prismatic section? Cannot find anything about it. I understand quartz is made of $SiO_4$ tethraedra, how does this relate to the pyramidal / prismatic sections?
Looking at the sketch below, two geometrically similar crystals (considering the pyramid-section) are presented. They both enjoy same equilibrium properties, same angles, etc. What caused one to appear instead of the other? Assuming growth starts by the vertex, what caused the transition from pyramidal to prismatic shape?

 A: It’s all about the atoms getting to the lowest energy position. If you put a layer marbles or beads in a shoe box and they are at random if you jiggle the box and tilt it back and forth you will see you can form a couple of types of 2D crystals. One where they are closed packed and you could draw little triangles between the centers of the marbles. Another where if you connected lines between the centers you would get a square lattice. These are like your habits, but in 2D instead of 3D
The same kind of thing happens in solution, or a melt or if you try to grow semiconductors in a vacuum chamber. Once you nucelate the crystal and have a small surface if the energetics are right the atom or molecule will find its lowest energy position and extend the crystal.  Kind of like how surface tension holds a droplet of water together there is a  quality called surface energy. Depending on the arrangement of the atoms in a particular crystal plane it may be easier for one crystal plane to grow faster than another plane.
When you evaporate the solvent in the solution slowly and you keep the solution very still you are giving the atoms more time to jiggle around and find the right place to extend the planes of the crystal.
Since the the planes are at different angles, eventually the planes intersect. That is how you come to an edge or a point.
In crystals like semiconductors, if an atom ends up in the wrong place, or if there is a defect. Sometimes you anneal the material by heating it up, not enough to melt it, but enough to jiggle the atoms around so there defect would go away.
In metals you have a lot of little crystals at various orientations (polycrystalline) and a metallurgist would call them grains and might anneal them at some temperature to make the grains bigger or smaller to change the properties of the metal.
So the concept of finding the lowest possible energy arrangement is an important one and used widely in physics and material science.
A: I don't think this interesting question has satisfyingly simple answer.
We know that crystal facets grow as molecules are are added or subtracted, e.g. see this old Royal Institution video and the Terrace Ledge Kink model, but as noted in the answer to How to make pyramid-shaped NaCl salt crystals: "There is no universal method to predict and to control crystal habits."
To illustrate the challenge, consider the the most familar hexagonal crystals: atmospheric water ice (commonly known as "snow").
When ice crystals form in air, their precise morphology depends on temperature, humidity, pressure, the presence of liquid water droplets, dust, or trace impurities, …. It is possible to identify over 120 snow crystal types, but the simplest snow flakes are hexagonal prisms with two basal facets (top and bottom) and six prism (side) facets.

The hexagon shape reflect the arrangement of water molecules within the crystal, and it is easy to see that the molecular forces for new water molecules arriving on a basal facet will likely be different from the side prism facets.  (With more complicated molecules, it often isn't obvious what are the likely molecular structures of different facets, so challenging measurements must be made.)

We typically need experiments to learn which facet's grow better under different conditions.   One study found that around -5 C the basal surfaces of ice crystals build up faster than the prism surfaces, producing columns, while around -15 C the reverse is true, producing flat plates.  The physics and chemistry of all this is not well understood and is an area of active research.
In general, trying to have a microscopic understanding of any specific crystal's morphology is tough, requiring heavy duty theory, supercomputers, and lots of fancy experimental equipment to make progress.
There is, however, lots of simpler science still to be investigated experimentally if your kids are into science fairs (and have lots of patience).

*

*Trace contaminants can be preferentially be absorbed onto some facets, causing them to grow faster or slower, so using tap or distilled or bottled water could make a difference. (For example, the aspect-ratio of ZnO crystals can be controlled by small organic molecules.)


*Do the aspect ratios change with growth temperature, concentration, or pH? (For example,  the relative growth rates of alum crystal 100 and 111 facets change quite a bit going from 0 to 5% supersaturation.)


*Are the crystals the same if they are grown on glass, different rocks, plastics, metals, …?  (For example, whether ZnO forms a pyramidal structure can depend on its growth substrate.)


*Does the crystal shape depend on whether you have a single crystal or multiple crystals growing in close proximity (as is common with crystal growing kits)?  Growing crystals remove nearby solute which is only slowly replaced by diffusion.
A bunch of neighbouring crystals deplete solutes near their bases and might grow faster at their tips where the concentration is higher. This could, however, be reversed by convection that can occur as the solution near the crystals becomes less dense because the solute molecules are absorbed and because the heat of crystallization warms the liquid. This convection can create a gradient with solute concentration decreasing with height, so now the bottom of the crystals may grow faster.
