We do not know how Virtualbox might have affected the performance. We check with the DVK on Win 10 and Win8 platform, we notice that it worked fine on both SS and HS mode. Can you once check using the same on a standard Windows platform (without Virtual box?).
Do you have any other USB/UVC camera? can you check if that works with Virtualbox? Is the Virtualbox hampering with the enumeration?
Will you be able to share a USB protocol trace recording the enumeration? This will tell us the details on the transactions on the USB bus (we can compare it with standard enumeration steps and see what is different).