I think the issue is, that if you have a thunderbolt over USB-C display, and plug it into a normal USB-C port on your laptop (say the latest macbook), it won't work, since the laptop doesn't support the thunderbolt signaling. So it's a new point of confusion - USB devices which are physically compatible but not electrically.

It's something that could possibly be solved through the markings on the devices etc and I still think doing thunderbolt over the USB-C port is great. Perhaps the display example could fall back to supporting a display-over-usb standard, it won't work as well but could help.

I don't think we have enough detail about Thunderbolt 3 yet to know whether devices will be able to detect that they are connected to a USB port and not a Thunderbolt port. If they can detect that, they may well be able to choose to degrade.

It seems like an ideal way for Apple to be able to ship a display which can be connected to both thunderbolt and USB-C hosts, without the user having to care.

Obviously some Thunderbolt devices (e.g. PCI-E chassis) wouldn't reasonably be able to degrade to USB, but a good proportion of them, would.

I very much hope this is what is going to happen!

The Type C spec also mentions a PCIe alternate mode. The main complication is to support both wrapper protocols given that thunderbolt pushes twice the bits per second over the same wire (if using active cabling)

Wrapper protocol?

The alt-mode stuff is a physical signal multiplexing, not a protocol encapsulation.

I don't think you're allowed to sell a device with a USB jack that won't work with USB, because the consortium has a trademark on it. Or perhaps you just can't advertise it as USB?

I'd hope they gave up on their own DP signalling and just use the type C one.