The surface of the Earth is tesselated into irregular tectonic plates that (mostly) move rigidly, with little internal deformation. Plate boundaries are geological faults (dislocations where the crust on one side of the fault slides past the other side), but there are also faults within the plate interiors, especially near the boundaries, that accommodate strain within the tectonic plates in 'deforming zones'. The rigid interiors of plates generally have strain rates at the nanostrain/year level, i.e. a few parts per billion or less per year. This means that the relative deformation (not plate motion) between say Florida and Denver is on the order of 1 mm/yr.
Plate motions are pretty constant through time, such that the velocities measured over 10 year scales with GPS and other geodetic techniques are quite comparable to velocities estimated over the past ~5 million years by looking at longer-term proxies such as the bar-code-like pattern of magnetic anomalies on the seafloor produced at divergent plate boundaries[1]. Within about 20 km of an active fault, or 200 km from a subduction zone, the instantaneous velocities have a lot of variation related to earthquakes, though.
The horizontal velocity of any point on the Earth's surface can be described as a rotation around another point on the Earth's surface, with some angular velocity; this is sort of a corollary or application of Euler's fixed point / rotation theorem[2]: the point is where the rotation axis intersects the Earth's surface (so there are two antipodal points, actually).
Because tectonic plates are mostly pretty rigid, the motion of plates on the Earth's surface can be modeled as a rotation around a single point, even when the plate is huge like the Pacific plate, i.e. almost a hemisphere. However as stated in the OP, there is no 'true' reference frame, so plate velocities are usually described relative to another plate as a reference frame, or just considering a pair of plates. We'd say that relative to the Pacific, the North American plate rotates around a point near Hudson's Bay, or equivalently, that the Pacific-North America rotation pole is near Hudson's Bay. To go from N. Am. relative to Pacific, to Pacific relative to N. Am, flip the sign on the angular velocity.
Therefore there is a bit of shared heritage between some aspects of graphics programming and tectonic geodesy, with respect to the mathematics of rotations, etc. However in practice the math is probably different (I'm not sure because I don't do graphics or game programming). For a good intro to the math used in this aspect of plate tectonics, see here[3].
In terms of the rates of the velocities, relative plate motion rates between adjacent plates are typically 1-10s of cm/yr. The fastest on earth is near Tonga and Vanuatu in the southwestern Pacific (north of NZ), at about 25 cm/yr.
[1]: https://divediscover.whoi.edu/mid-ocean-ridges/magnetics-pol...
[2]: https://en.wikipedia.org/wiki/Euler%27s_rotation_theorem
[3]: https://websites.pmc.ucsc.edu/~thorne/EART118/Lecture_PDF/le...