Lossless PCM (compressed or not) formats are all well know. They are easy to edit, transport, and decode. There are plenty of theories that accurately describe what to expect from a specific bit depth and sampling rate.
DSD is an odd one. It's a digital format that attempts to act like analog. It has a slew rate (i.e. it can't go from silence to maximum volume instantly). There are no easy ways to manipulate the data. Sony was pushing the format on it's recording studios. When the engineers asked for tools to perform basic mixing tasks Sony delayed, and delayed, and eventually shipped tools to convert DSD to PCM and back. So the work could be done in PCM. It is likely that very few SACDs are direct analog to DSD conversions. Instead they made a trip by high resolution PCM. PCM to DSD is a lossy process, there are valid PCM signals that can not be stored in DSD.
The math for estimating the limitations of DSD is not as clear as it is for PCM (with the work from Nyquist, et al.), but it does seem that the signal to noise ratio at the high frequency end (still in the human hearing range, though) is very poor. Worse than a well mastered CD in fact. Now, when it comes to storing bass information DSD far exceeds everything else. But for quickly changing signals it doesn't keep up.
Here's more: http://sound.westhost.com/cd-sacd-dvda.htm