I seriously doubt the reverse sublimation theory. This would not explain the shapes. Reverse sublimation, if it’s even possible with water vapor, would tend to be more evenly distributed so that even if you did detect peaks, they would tend to be humps - not cones, rods, or wedges.
Here’s my working theory, and it’s based on an observation I saw in a friends’ freezer a couple of years back:
The defrost cycle (if not optimized) continually creates a pattern of thawing, condensation, and refreezing. Depending on conditions, this can create stalatites (icicles) and/or stalagmites (spikecicles). Either way, these formations will have ample opportunity to grow. They may tend to congregate in areas of higher water vapor (i.e. over the top of an ice tray).
If they grow from the top down, eventually, during one of the defrost cycles they may grow heavy enough to lose their grip and fall. If they land on an ice cube, the small amount of liquid water on the surface can instantly freeze - sticking the icicle in a mostly upright position (similar principle to sticking a cold ice cube to your tongue). In fact, I have seen such tiny icicles growing over the ice trays in my aforementioned friend’s freezer. BTW, once the icicle falls, depending on a lot of different factors, it can reflow into a number of different shapes or develop a slant, etc… Another factor in support of this theory are the wedge-like formations in some of the pictures. These look remarkably like fallen “ice curtains”. The only problem that I see with this theory is that you would expect to find a few fallen icicles laying on their sides and some still hanging from above the ice trays. These phenomena have not been described (which doesn’t necessarily mean it’s not occuring).
Alternatively, if the defrost cycle is just slightly warmer, the condensation might fall and the small water droplets would refreeze fairly instantly forming stalagmite like structures. These might tend to be mostly vertical, but defrost cycles and air currents from blowers might make them tend to lean once they reach a certain height. Come to think of it, a leaning structure, under the right conditions might take on the shape of the wedges depicted. One way to check out this theory would be to look for smooth bumps directly above the ice trays. This would be indicative of condensation droplets freezing and thawing.
I’m sure there are a lot of holes in this theory, but I haven’t seen anything in either thread to necessarily contradict it…