First time I've been accused of lacking respect for a software application :-)
I've made a bit of headway. When I use my Mackie Blackjack interface (ASIO) everything is fine and dandy. I can put everything to sleep with Sonar open and when it wakes back up it'll resume recording no problem. Being usb powered, the Mackie is getting a hard reset each time, which may be helping.
With the Soundblaster (ASIO) [I know a lot of people don't want anything to do with Creative, but for the moment I'm giving them the benefit of the doubt, until proven otherwise] it actually worked a couple of times during my trials, though sometimes at reduced volume. It hadn't worked at all before. Windows and at least browsers apparently have no problems with the card. I don't yet know what happens to the PCIe bus and cards plugged into it during sleep.
I have to study up a bit on how audio is handled and routed internally to Windows. I'm missing some big pieces of the puzzle which makes testing kind of vague.
I believe software should be solid enough to work as is without having to baby it or tinker endlessly with it. It shouldn't be possible to break it, at least not easily. My benchmark is that a little old lady should be able to use an application. I've worked as a software designer and programmer on large programs in another life so I have an idea of the complexity involved. I feel sorry for the little old ladies of today.