With HID enabled, too many threads running and tanking the USB IP due to lower bandwidth? However, thought that the XCORE-200 processors no longer have the min bandwidth requirement for USB but my memory may be fading here.random thought...
Does the HID leg of the IP work ok? Is the HID code factory fresh from XMOS? If not, consider to import a proven code run for the HID support from XMOS and test again.
Check the resources being used by your IP to rule out if bandwidth is a factor.