Andrew, I really would like to know the pluses.
There's an inherent long-term plus in the format being publicly documented, and also in the file having one or more updated JPEG previews that other opps can display or output. So for instance, I could abandon LR and I would be able to see the adjusted image in Aperture's preview mode, in browsers like PhotoMechanic, or in cataloguing apps like Expression Media. These apps can read and display the updated preview JPEG that's in the DNG, do so extremely quickly, and without needing knowledge of the Adobe/LR adjustment parameters (one had a good laugh at the earlier comment about "normal cataloguing apps are as good w/ non .DNG raws as they are w/ .DNG raws"). How that benefits you will vary, but for example there are people who use Bridge or even LR for processing, but catalogue their work with programs like Expression or Portfolio - and they see the adjusted DNG, pretty handy if you've had to make major corrections.
Another set of benefits is that the metadata is embedded in a file rather than in a sidecar xmp file. Many more 3rd party apps read metadata that's in the file than read it from sidecars (which can easily go missing). And that's even more true of writing metadata. So for example I might want to tag images from a trip with GPS co-ordinates. Apps like HoudahGeo on Mac or Geosetter on Windows will write the GPS EXIF metadata back to the original so it's available in Lightroom or other apps. Alternatively, I might change pictures' captions or star ratings in Expression Media or PhotoMechanic, and again want that metadata attached to the file itself, and therefore into an TIFs I produce from it, or JPEGs I put on Flickr or other online service. What do you reckon is safer - writing that EXIF and IPTC to a publicly documented format or to a proprietary one?
So apart from the archival value (objections tend to be along the lines of "hey, we'll surely have time to fix the roof when we see the storm clouds"), the key pluses are embedded previews and safely-embedded metadata, and how you can then take advantage of them in diverse workflows.
John