From my point of view the main limitation of the Go for games isn't 3 DoF or the limited hardware, but the controller. With only a trigger, a (clickable) touchpad and a back button, games often struggle to offer a good method of control.

Even just adding a shoulder bumper or the like might help in some cases (for example in Pinball FX2), but a second Go controller with a thumb stick should do the trick in adding the right amount of interactivity, achieving most of what a game controller can do and some things it can't (such as separate interaction for both hands).

An ideal second controller would look like the regular Go controller, but offer a thumb stick in place of the touchpad. A plastic bridge that the controllers slide into could help join them into a single controller, for games that may be more comfortable with a classic style game controller.

It's not a trivial challenge to add this, but doesn't seem too complex, either. I'm not a hardware designer, but (looking at a controller teardown) it seems to me that a thumb stick would take some vertical space currently used for the board. If a redesigned board could fit around that, great, but it not, then the controller might have to be a tad bigger. Not a big issue, but would kill symmetry. Then again, if the Go's touchpad controller is modified to add a shoulder bumper, maybe that would return symmetry back.

Another challenge is if we want the second controller to have similar directional tracking as the first (which I think is highly desirable), allowing for two handed games, like Beat Saber, using sword and shield and so on. (Obviously we don't get spatial tracking, but even with its 3 DoF limitation the Go controller has already proved adequate for this kind of control.)

With two different controllers, we can easily assign one to be the left controller and the other to be the right controller. Both user and headset will be easily able to tell which is which, so we could extend the current body model easily, something that won't be trivial if just using two of the same controller.

Then we need to take care of drift. The current method of recentering seems fine to me, and I think can be extended to a second controller. The user would simply hold the two controllers while resetting. I think that most users would be able to adapt to this easily (I tried holding two controllers like this and it felt fine to me), and since the second controller would be optional, apps that don't ask for its use will not require it to be centred, and the OS could even turns it off (which should help battery life).