This speculation as to the process of crystallofolia is based on my examination of the current literature on freezing in plants (especially trees) and plant anatomy. None of this literature deals with crystallofolia and extremely little has been written concerning herbaceous perennials similar to plants known to exhibit crystallofolia. I have thus been forced to put together a somewhat eclectic set of hypotheses and claims in the scientific literature and in online botanical presentations, seeking a best match of apparent possibilities to account for the wide range of phenomena for which I have direct knowledge. At a late stage I modified my views after consulting with a specialist in plant anatomy (James Mauseth, author of Plant Anatomy), and am much in his debt — although he did not examine my written scenario and is not responsible for possible errors in the views I present here.
The areas of stem structure relevant to my discussion are (as shown below):
Thus the initial opening occurs above the base, which leads to dehydration of the stem above that point. Subsequent ruptures move down the stem. At the very bottom of a fully opened stem, exposed xylem rays and secondary xylem are still capable of raising limited amounts of sap  from the roots to the opening — perhaps an indication of pressure from the roots in the absence of intrastem pressure.
The epidermis rupture is primarily a longitudinal fissure, following the organized structure of fibers and tissue strands. For the most part, the cortex remains bonded to the epidermis and these together tend to detach from the woody secondary xylem.
Section of V. virginica after frost formation.
One ray (center of red frame) may have produced ice.
Detail of ray area on the left.
click to enlarge)
3-dimensional schematization of vessel / ray relationship.
(Highly oversimplified; not to scale)
In short, the sap rises axially in vessels to a point where it then moves laterally outward via horizontal series of rays cells. With the splitting away of the cortex, this row of ray cells now terminates in the cold air, each terminus functioning as a small 'spigot,' and jointly with others forming a vertical band of ice crystals, a ribbon. Close inspection of ice formations, large and small, show that they have all been formed in this manner.
According to my observations the number of ribands vary from one to five. All of them issue from the stem in vertical or longitudinal lines, which are not always symmetrically disposed around the axis. [p. 22]Le Conte, however, was eager to establish a connection with somewhat similar ice formations in soil, and probably could not have been aware of a number of essential differences; i.e., the complex anatomy of tissues still living at the time of the formations.
The porous pith furnishes a constant supply of warm water from the earth, while the wedge-shaped medullary rays secure the mechanical conditions necessary for the development of a projectile force in the proper direction. In proof of this, it may be remarked that the medullary rays are very conspicuous in the Pluchea, and when the stalk is split by the freezing of the water within, the fissure is observed to follow their course. The development is arrested when the pith becomes frozen, for the obvious reason that the consequent splitting of the stem destroys the arrangement of the resisting tubes. [p.33]
The reason why the phenomenon is manifested only in certain kinds of plants, probably arises from several peculiarities in their physical condition. They must be porous to furnish an abundant supply of fluid. They must be herbaceous and annual to secure medullary rays of sufficient size and openness, and, it is probable that all vital action must have ceased, in order that the fluid which is released from the soil may be unmixed with the proper juices of the plants; a mixture which would interfere with congelation. [p. 33]
One property common to our species is the formation of buds and/or leaves for the coming season, beginning already in December. This was noted already in Scientific American 36, 1877, by J. Stauffer:
I took up a number of the [Cunila origanoides (common dittany)] plants, and always found a vigorous scaly root bud, undergoing development at this early season under ground, to produce a new stem the following spring.
This was most striking with Pluchea, which in mid January already had produced rather large leaves at its base.
P. odorata, Jan. 20, 2007, day of 18° F frost.
The right photo is from the red frame on the left.
Base of a stem cut 30 mm above the ground (shown with crystals immediately below, later the same day)
The area right at the soil level has been cleared away, exposing an enlarged base (red line).
The epidermis was still intact at the base (green line).
A section of the epidermis was torn off, removing a bud – yellow arrow.
Green leaves are from this plant.
[For additional pictures of this cut stem over a month period]
Pressure to initiate crystallofolia at the very base is assumed to arise from all living cells near the ground, in the leaves as well as the cortex, via extracellular freezing. Unlike the events which cause the stem to rupture, these freeze–thaw events are part of the plants normal strategies for surviving low temperatures — water moves out of the cell, increasing the concentration of solutes in the cell, thus lowering its freezing point. With warmer temperature the water returns to the cell.  The resulting pressure is assumed to impact the xylem conduits throughout the plant, including, via the roots, the secondary xylem vessels of the stem. Water is thus pushed up from the roots into the dead stem vessels, and into the rays through the pits in the vessel walls.
Roots are examined on a separate page.
Earlier exposure (with epidermis opening) is assumed to have led to fatal freezing of the ray cells, enhancing the intercellular flow and probably enlarging it. In affect a direct (including a right–angle turn) pathway from the roots to the edge of the woody xylem has been created, and water pushed up from below is forced out. Subsequent freeze–thaw events enlarge the ray conduits even more, enlarging both the diameter of the output 'threads' and the length of the ice bands. Note the following photos with repeat formations:
P. odorata, Jan. 20, 2007, day of 18° F frost.
Details of ice crystals after repeat formations with no stem rupture.
Direct root pressure would also seem to be a factor for transporting water to the dead xylem, along with normal flow to the winter buds and new leaves at the base. But this alone doesn't seem sufficient to initiate the crystallofolia. Le Conte noted that some root damage didn't suffice to stop the process, only reducing its scope:
Plants which were torn up and transplanted in a vase of moist earth, in a flowergarden, exhibited the same phenomenon, although much less strikingly than when left in situ. [p. 24]
D. T. MacDougal (1893) claims to present evidence to the contrary, arguing against root pressure:
It appears that the water is taken up by the simple saturation of the roots from the charged soil, without the intervention of the special activity of the root hairs, as is shown by the fact that plants dug up and replanted, which would destroy the larger number of the root hairs, still formed crystals as usual. Then root pressure must be wanting as well as osmotic activity in plants at this stage. [251–252]The facts as stated by him don't support his claim. For one thing, we are not told the size of the plants and whether they had previously exhibited ice formations.