What works in practice?
On most desktop PCs, USB 2.0 ports on the front/top of the case are connected to the motherboard via a cable terminated with standard 0.1" connectors. Notice there is about 1-2cm of exposed unshielded wire, and 0.1" headers are not really "high-speed"...
![enter image description here](https://cdn.statically.io/img/i.sstatic.net/hodpi.png)
The USB connector shield can make contact with the metal case, for shield continuity, but it is not always the case. Sometimes it's all plastic everywhere and it still works. I'm not sure about EMI certification though...
![enter image description here](https://cdn.statically.io/img/i.sstatic.net/N2qAJ.png)
I don't think your pogo pins will be worse than this abomination... And there are zillions of motherboards working without trouble using just that... Real-world tested!
I'd rather use the USB ports at the back of the PC as these are directly mounted on the motherboard, in order to avoid culumative signal degradation from the 0.1" headers and wires on top of the pogo pins.
If this does not work I'd try to solder the pogo pins on a USB male connector and plug that into a hub directly without using a cable. A few cm long USB connection (between the hub chip and your board) without extra cable length should be short enough to reduce transmission line effects substantially.
I'd also program the microcontroller with a USB2 bulk echo program, then use libusb on the PC to transfer lots of data back and forth, let it run for a few hours, check if it works, how fast it is, add some extender cables to degrade the signal to see what margin you have, etc.
Note the loose wires are no longer an option for USB 3.0... but it still uses 0.1" headers!
![enter image description here](https://cdn.statically.io/img/i.sstatic.net/mm8kC.png)