NOVA seems to be a way to help make mods more version-independent with respect to vanilla updates. All in all, it seems to be a good step in the right direction, since it'll give modders less backend stuff they need to worry about, thus allowing them to focus more on content, bugfixes, and the like for their own content. Forge is still quite beholden to Minecraft's update schedule and the ways in which it can, and at times will, break things rather badly. I applaud it.
To put it into real-world terms, think of it like this:
As it is now, when Reika goes to update RotaryCraft for a new vanilla version, he has to take into account not only the various changes that inevitably pop up with Forge, but also the massive changes under the hood in vanilla code, like the massive changes to rendering in 1.8. Accounting for these changes takes away time from Reika's schedule that could otherwise be spent on adding content to RotaryCraft, improving its functionality, and generally making it a better mod than it already is.
With the abstraction layer provided by NOVA, and operating under my understanding of what that means, Reika would now have fewer "behind the scenes" changes to worry about since NOVA would take care of that internally. NOVA's wrapper classes and methods encapsulate Forge's (and perhaps vanilla's, to an extent) code, and delegates between them and RotaryCraft; it speaks the ever-evolving language of Forge to Forge, but when speaking to RotaryCraft, it will be using the same words and syntax every time, making it easier for RotaryCraft to understand. This frees up time Reika would otherwise be spending on accounting for Forge's various changes, allowing him to get more done with respect to his mod's actual content.
Again, this is all working off my own, personal understanding of how NOVA works, so take it with a grain of salt. I, like anyone, am totally capable of being utterly, completely wrong.