Of course, the lack of data integrity may be the worst failing of the DWG "technology", but that isn't the only one.
We can look at the overall "DWG technology" (i.e., the code that writes data to and reads data from DWG files) as the persistent storage mechanism for our data. In other words, the "DWG technology" is basically a form of relational database. However, the entire DWG structure was designed around a single-user paradigm. Only one user can be editing a DWG file at one time. It is not multi-user, like most modern database technology.
Once upon a time, that was fine. The DWG technology did what it needed to. And in fact, when Autocad first started, the computers could not have handled the demands of a rigorous solution to this problem. But that has changed, and the computers that fit in our pockets can outperform the supercomputers of thirty years ago, and are perfectly capable of the task. Also, with the growing number of giant projects, there is a large demand for multi-user software. And this is not a new turn of events; it has been growing for years now. The lack of multi-user access to data was a common complaint with Land Desktop, so it's definitely something Autodesk has been aware of for some time.
Ideally, Autodesk would have addressed the issue at its core, and made Autocad truly a multi-user product. And ideally, they would have started addressing this problem in a serious manner as soon as it became obvious. But that would mean replacing the DWG file with some other technology. So instead, what they decided to do was implement Vault, and let Vault "oversee" a whole collection of DWG files. In essence, they tried to "back door" their way into multi-user technology by utilizing their old, single-user DWG technology. Then, by simply using a lot of DWG files, they created a simulation of a multi-user environment.
The biggest problem with this, as mentioned earlier, is that the DWG technology itself is unreliable, and corrupts data too easily. So, they took a piece of technology that didn't really work in the first place, and made it the core "building block" for their new, multi-user attempt. This fails in multiple ways. First, the underlying technology STILL does not support multi-user usage. Second, the underlying technology is unreliable. Third, the technology they are using to tie it all together (the Vault) was designed for another purpose entirely, and is ill-suited for its demands as "overseer of the multi-verse". The whole design is enough to make a software architect like myself shudder with fear.
The best construction analogy I can come up with is that they knew the foundation was bad, so they decided "well, let's just make the walls really thick...". In reality, all that does is make the walls more expensive, without fixing the bad foundation. So when the foundation finally fails, the failure is more spectacular (and costly). Unfortunately, that's the most-likely result if Autodesk continues along the path they are currently following.
