From what i understand about DSD , it is not necessarily the file type / sample rate itself that offers the sonic benefits, it is the performance of the converters and the way in which they operate (in the 1bit mode and the lack of filters, etc, im not an expert). That being said theoretically a file recorded in DSD and then converted to PCM should sound better than one converted using PCM. I have read peoples opinions though saying they couldnt hear any difference in this scenario, but in theory it should sound better. Assuming the conversion is better, it would be of benefit for people processing mixes, stems, or tracks through outboard gear to convert the pcm files to dsd, transfer to flash card, and playback that converted dsd file via the da3000, and capture using a 2nd da3000 (as far as im aware you cant playback and record dsd simultaneously on one unit). It would be nice to have synchronization features in sonar to play nice with the da-3000 or multiple da-3000's. One possible setup would be to have multiple dsd decks connected with a summing mixer, and another dsd deck to capture the summing mixer output. You could stem out your sonar stems in dsd and transfer them to the dsd decks, while having a dsd deck for capturing your hardware synths (thus avoiding ever converting hardware synths to pcm during mix) , or you could sequence the hardware and just blend it in the summing mixer during mixdown and save having to track it to its own unit prior to mixdown. The most basic setup would be to just have 2 da3000s used for mixbus processing and possibly minimizing conversion loss from the round trip. Next step up would be to have say 3 da3000s for stems connected with summing mixer, and a 4th to capture. If the synch of the decks themselves, as well as the synch between grid/sequencer synch between sonar , and the dsd3000 aren't flawless then just basic 2track mix processing loopback chain would be all i would want to mess with.
While not an easy workflow, if minimizing conversion losses during the production process translates a benefit to the end result, it might be worth it to go through the hassle of tracking , transfering , converting the dsd files back and forth. I don't own a da3000 or any other dsd recorder so i can't say yet if you do get a sonic benefit from tracking dsd and converting, but if there is ,then people might want to track things like vocals, etc on the da3000 and import and convert to pcm. If sonar could come up with some form of communication and automatic file transfer from the da3000 it might be really cool. For example you hit record in sonar and are doing a vocal punch at 1:30 on the timeline, and sonar sends the record message to the da3000 where the audio gets recorded, and then upon hitting stop in sonar, it automatically pulls the dsd file from the da3000 over usb and imports and converts in to sonar placing it correctly on the timeline. Or better yet just make an dsd audio interface that works directly with sonar, or somehow add this function to the da3000. I know that with this interface you still wouldnt be able to work in dsd once inside sonar, but if the converters do in fact perform better , it might be worth it ( even with the degraded workflow ie: waiting for files to convert after every take)