Yes. The problem is security, because USB has traditionally been restricted to relatively simple devices that have not required much in the way of access to host resources; but PCIe devices require direct access to both memory (DMA) and the PCIe bus, which are extremely highly privileged activities. So USB's current model of "plug it in and it just shows up in My Computer" absolutely cannot work for PCIe-over-USB devices; there will need to be some sort of "Do you want to allow this device to connect?" prompt for such devices, and how that is implemented is as yet unclear (is/should it be an inherent OS function, or should USB firmware control it a la Thunderbolt's implementation?).
Security is also a factor in the actual data transfers because PCIe-over-USB devices will be operating in a limited trust environment, versus ordinary PCIe slot-connected devices that operate with full trust. That means that every single operation attempted by a PCIe-over-USB device will need to be verified by the USB controller to which that device is connected, to ensure that the device is actually the device it claims to be and that it is allowed to do what it is trying to do, and probably a whole number of other things I can't think of. This introduces inherent latency that simply cannot be overcome by anything.
Now, Thunderbolt has already encountered these issues and mostly solved them, and of course USB-IF has the TB spec now, but... TB was designed from the ground up with direct access to privileged host resources in mind, while USB was not, so how much of the TB implementation can or can't be reused for USB is unknown.
tl;dr when security enters the picture, nothing is ever simple.
