One reason to improve tile support sooner rather than later: The sooner people can get to work on better tiles, the more tilesets will be created in the long run.
I know it isn't a problem for people here, but I've seen (on many forums) people saying that they'd play Dwarf Fortress if only it had a somewhat 'better' graphical presentation. Just allowing tilesets to change the appearance of map features without changing the alphabet would be a huge improvement. Allowing tiles to be specifically set in the raw files (so you can have a specific tile for monster type XYZ) would be even better.
Yes, tiles will never capture absolutely every bit of detail, but it has the potential to be a big improvement for some players. Not everyone likes ASCII. That doesn't make them lowly cretins unworthy of consideration; it means different people have different tastes.
(Although, I do have to say: It is also possible to have a graphical presentation for Dwarf Fortress that is superior in terms of information conveyed than the ASCII one we have now. The limited ability to see things on lower Z-levels, the fact that clouds completely obscure anything in the square, and -- when lighting is in -- giving a rough sense of light levels/color, where light is originating from, and so on are all relatively simple things to convey with a pseudo-tileset. Perhaps this should wait until then, but saying "Oh, ASCII is just fine" avoids the issue that, yes, there is the potential for genuine advantages in a more detailed system. I don't care about the rest, but being able to actually see large numbers of Z-levels effectively would be a huge improvement for me.)