Out of curiosity, is there any criteria for which cameras have a second buffer and which do not?
Generally the ones with more physical RAM, which also tend to be the more expensive ones.
If you are thinking about optimizing DNG, you should figure out what gains are available before worrying too much about possible implementations.
Writing a full size 4104x3048 17.89 MB CHDK raw (full sensor, all one write, no reversing) on my D10 takes ~3 sec on a slow card = 6 MB/s. The active area is 4072x3030 = 17.65 MB. So if you removed all the inactive area with no overhead, you would save writing 0.24 MB = 0.04 sec.
I would certainly take this if it were free, but in reality there would be some runtime cost to set it up, and also code complexity. It's not unlikely that the former would be a significant fraction of the savings.
Moving to a class 10 card would give me a much better gain, and the time saving available from skipping the inactive area would be correspondingly smaller.
A possibly better approach would be to use lossless compression on the data. The cameras that support native raw already do this, so there must be some way to do it quickly on the available hardware. In theory, this could be done in parallel with writing, as we do for reversing. Lossless compression should typically get you something like 15% reduction in data size, which should result in a corresponding saving in write time. However, even if it's theoretically possible, cameras without two raw buffers could still be problematic.
Another factor is that the masked area can actually be useful, because they give a measure of the black level for the specific exposure. The MaskedAreas DNG tag can be used to explicitly support this, though it would need to be set up for each camera.