Network interfaces blur hardware and software issues badly. Network hardware is supposed to do TCP offloading and checksumming. Cheap (or rushed to market) network stuff wants the OS to do this. Usually, the hardware, or the OS doesn’t do something Internet Protocols fundamentally expect, even when that OS is Linux or BSD. Add to that multiple translations from CPU to PCIe to USB, and you have a recipe for disaster. This breaks protocols as basic as HTTP all the time.
Most users just don’t notice the occasional missed webpage reload.
Gaming/VR is realtime, which is not ‘most users’. Neither are server admins who hold onto PCI-X (yes, PCI-X) 15+ year old network cards because of this.
In this case, Pimax might be able to add a reasonable workaround for whatever is going on, or at least a warning message about USB WiFi.
When building VR workstations though, everyone needs to remember that we are building realtime systems here.
Would you feel comfortable designing a 3D printer to send STEP/DIR signals to the motors over a USB WiFi dongle with an unknown chipset, connected to a USB controller with an unknown chipset, shared with other unknown USB devices, with all of these things having unknown buffers and error rates at every layer, and protocols that just drop connections when some of these failures happen?
I don’t. But with a conservative approach and some testing, I am planning to directly control 3D printers by commodity PC hardware. When we build VR workstations, we must do the same.
The OP’s issue report should be a sobering reminder to us of the reality that we are putting humans in flight simulators connected to motion control hardware, all driven by many layers of fundamentally not-realtime systems.
So, please, buy decent hardware.
EDIT: Offtopic, I am seriously considering selling ready-to-go VR PCs with some very innovative features.