Theory - One time calibration possibility.

[uq1]

This would still require initial calibration.

SteamVR has a room center marker, as does every other play space method (I think?). If not, encapsulate the spaces and find the center of the box (actually this would probably be better anyway, Pico doesn't make it's center the center, but instead it's where you turn the damn thing on, grrrrr). If the play space systems do not have a rotation, assign one when first calibrating for each. It's a bit like having a virtual tracker in each play space system.

You now have 2 positions, and 2 constant rotations that can be continually recalibrated from. No need for a tracker on the HMD any more, and if the play space moves a little (eg: because of inside-out lighting issues or whatever), the center of each play space is still the center... and as far as I am aware, the scales and rotations never change (unless someone redraws the boundary).

Not sure if I am missing something here, but wouldn't this work, and permanently remove all drift?

Anyway, just a thought, not sure if you have considered this.