They do. Cable modems are provisioned with multiple service flows much in the same way cable digital phone service works. I've seen this setup in a lab being run from one CMTS. The modem has a service flow for Internet (say 50Mbit/sec) and another for the eSTB (say 30Mbit/sec) which can handle the IP VOD and IP multi-cast video. It is possible to have multiple CMTSs for this however both Cisco and Motorola are investing in VDOC (video-over-DOCSIS) scale solutions that allow you to put all your eggs into one huge basket. One cable modem / gateway will ultimately be feeding every cable service in your home and some portion of it will be provisioned for plain ole' Internet traffic while the rest of it will be allocated for MSO services. From what I can tell this is not how Comcast is delivering IP VOD to the Xbox today. I would expect them to make the transition fairly quickly though. This stuff has been up and running in labs for the last 5+ years. MSOs are just very slow to adopt new technology.
I think it's technically impossible for Comcast to get video to the Xbox w/o using the IP connection provided by the customer's HSI service. I would say, so long as it's using the same IP the customer would use to get on the Internet with, that it is using HSI and should be subject to whatever limitations they place on HSI.