If you have any sizeable stylesheet these days, you should use a CSS preprocessor. They're awesome, but they come with a few cons. The pros still definitely out-weigh the cons. Use the preprocessor. The cons are just things you'll want to look out for to make your experience with CSS preprocessors better.
CSS Preprocessors are Popular
Chris Coyier recently took a poll on the usage of CSS preprocessors. About half use one of the preprocessor options available to them. Of the languages used, LESS is the most popular. (Stylus is my current favorite. I like the HAML-style syntax.)
This means that half of the people who write CSS aren't, in fact, writing CSS. They're writing in some language that transpiles to CSS. So, half of these developers are getting the pros, but there's a bit of baggage too.
Cons of a CSS Preprocessor
File Size is Deceiving
When you're writing in a higher-level styling language, it's often much smaller than the generated CSS will be. Thus, you could be deceiving yourself thinking that your resulting CSS will be just as a small as what you're editing. Why?
Hiearchy will be flattened in the resulting CSS. And the tree of selectors will be duplicated for every selector.
You can loop in a preprocessor, generating n items where you only see one. All those items will be listed in the final, static CSS file.
In some of these languages, you don't have to write vendor prefixes. This is a fantastic feature, but it makes you tend to forget that one attribut will expand into 5 or 6 in the resulting CSS. (Stylus' nib is great for this, btw.)
Maybe not a big deal. Remove an image from your site, and you've made up the difference in bits on the wire.
Because you usually won't be deploying your .less or .styl file, you'll need to have a point in your development lifecycle where the .css gets generated. This can potentially slow you down, but there are great tools like Grunt or CodeKit to make this quite automagic. These tools have good, logical defaults, but still require configuration for your specific project.
Hard to Go Back
After working with Stylus or another language for a while, you'll meet a project that is straight CSS, and the RoI on conversion to a preprocessor is small. So, you'll be back in CSS land. This will be hard. You'll realize how nice you've had it. The mechanisms you've grown to rely on won't be there.
Variety of Syntax
Each CSS Processor's syntax is different. The features largely overlap, but they're implemented differently. It's a polyglot world, so it's no big deal. Still, it can be confusing switching between projects that have different styling syntax. You'll write your styles twice after you realize that Stylus syntax doesn't work in a LESS file.
You need to get everyone on your team on board when using a CSS preprocessor. If someone starts editing a .css file, their changes will be overwritten. Minify the generated .css. That should make it less savory to edit. It'll be a deterrent, but I can almost guarantee someone will edit it anyway. :) Make the stylesheet workflow very clear.
The Native Future
We really only have these CSS preprocessors because CSS is too strict, flat, unflexible, and verbose for the amazing things being written for the web today. Wouldn't it be nice if these styling languages or some new, similar version of CSS was available natively in the browser?
Meanwhile, we'll use these tools. What have you found your greatest challenge to be using a CSS preprocessor?