Let me see if I can answer some of your questions -
1. Polarity - It has to be in software, reversing it in hardware means all your ones become zeros and zeros become ones. This isn't analog! Conceivably, a bit stream could be designed for that but to what end? If it is done in hardware, then it takes some silicon to covert the phase, so SW is usually the way to go.
2. Clocking - 10m? I am not sure what you mean by 10m. I believe the question you need to ask is about jitter.
3. Galvanic Isolation - You are correct, galvanic isolation is the way to go. The problem is, it is relatively expensive. If they can achieve similar results in SW, then that is the direction they are forced to go. Note that the same thing applies to balanced inputs and outputs in the analog domain. Specs love to quote CMRR but it is CMB on the output device that dictates how good the CM rejection is because excellent CMB is much harder to achieve. Galvanic isolation on the output is really the way to go, yet it is a cost issue, and many, if not most, audiophiles would ignorantly question a transformer in their signal path.
4. Class D - It is still a young and evolving technology, I do believe it is much too soon to expect standardized clock frequencies or power supply topologies. Designers need to have the flexibility to improve their design as they see fit.