Seems like it'd be great in theory.... The docs are giving me trouble so far though. Differed isn't working for class selectors.
OpenSCAD released an update. Version 2015.03 of OpenSCAD offers a lot of new functionality that concerns relativity.scad. Chief among the new functionality are list comprehensions and the concat() function. I've really wanted these functions for some time. For reasons I won't get into, they greatly simplify the backend to relativity.
You might not notice it, but there's been some sweeping changes within the St. Patty's release of relativity.scad, most of which is meant to take advantage of this new functionality. Most changes are confined to the string processing library, String Theory. Relativity uses this library to implement selectors like "foo, not(bar)". Unless you use the string processing functions that are packaged with relativity.scad, You should not notice anything. If you do notice anything, make sure to file a bug report. If you do use the string processing functions, check the link above for news on the update.
Due to the sweeping nature of these changes, the St. Patty's release is not backwards compatible with OpenSCAD version 2014.03. If you still use version 2014.03, you should keep the version you're currently using. You can still find the latest 2014.03 compatible version here, in case you lose it.
Also, due to some unforeseen behavior in the newest version of OpenSCAD, there's also some foreward compatibility issues with older versions of relativity. Again, these issues only occur when using older versions of relativity with the newer versions of OpenSCAD. I'm only one man, and given the extensive changes within the string processing library I've decided not to patch the old version of relativity.scad to operate with the new version of OpenSCAD. Doing so allows me to focus on the future of the library.
TL;DR: If you're going to update OpenSCAD, make sure to update relativity.scad.