I think the base problem comes from the fact snow tiles, currently, are almost dead tiles, which reinforces the problems linked to map contingency (when you're dealt bad tiles in your starting area). They are very different from desert tiles, who have a lot of interesting spots (hills, oasis, along rivers).
As pineappledan said, snow tiles in Civ V are more akin to northern Greenland, totally unhabitable areas of the earth, with no yield at all from any tile (no even hills). Hence why a lot of people want to make unique improvements able to be built on snow tiles, despite how incoherent it ends up graphically : they are one of the rare means to counter this contingency problem.
To me, the solution is different, and in several steps :
- First, rename tiles so that they have the correct names : I don't know if you're aware, but the name tundra designates a biome where almost no tree grows, and is basically what snow tiles should be meant to represent ; their actual name should be Taiga, and Snow tiles should be renamed Tundra (see images below).
- After doing this, we should give the new Tundra tiles the means to be actually useful : through a pantheon, more mineral resources and the fact you can have yields on hills, I think we could transform these Tundra tiles into biomes where correct investments could reward you with science, production and resources.
- Finally, Ice tiles, which currently are once again dead tiles (with problematic impacts on City-States and players spawning just next to them), should also have some utility, but only in the endgame : give worker and special units the ability to walk on ice and build some scientific improvements (like a research station, whose 3d model already exists in the modding community).
In the end, making things this way would alleviate the gameplay problem currently posed by Snow tiles by giving some precise means to compensate for the challenge they pose. This would allow us to remove almost all unique improvements from Snow tiles, and restore a bit of graphical logic.