There are two basic problems in the current reference frames. One is discarding the underlying geophysical processes, especially in the Earth mantle, and the other is instability or evolution of reference frames in time. Both of these problems become crucial when rate of the deformation is very low.
In reality, the stable plate interior deforms very slowly in a complex way due to phenomena such as GIA and other mantle-scale processes associated with the heterogeneous lithosphere which is occasionally hosted to large intra-plate earthquakes. Until recently, such slow intra-plate processes have been ignored in the underlying models of reference frames (Blewitt et al., 2005). Moreover, the GIA is known to produce greater intra-plate deformations than plate tectonics across a large portion of the North American plate, and it would need to be considered in the reference frame. Horizontal vectors, however, might be considerably biased by differing reference system rotation-rate vectors (Henton et al., 2006).
In principle, the scale in ITRF2000 is derived from observations using very long-baseline interferometry (VLBI) and satellite and lunar laser ranging (SLR/LLR), while only VLBI has been used to determine the scale in ITRF2005 (Altamimi et al., 2007). Because SLR is the only method able to sense the Earth centre of mass, with accuracy comparable to the positioning accuracies achievable from GPS or VLBI, the Earth centre of mass is entirely derived from SLR in both the ITRF2000 and the ITRF2005. These considerations about scale and Earth centre of mass are also valid for possible changes in scale, or stability in the determination of the centre of mass of the Earth in the different realizations of ITRF. For example, instability in the centre of mass would map at a 1:1 ratio into derived station velocities, while a possible instability in scale at the 0.1 ppb yr-1 could contaminate the derived vertical velocity by about 0.6 mm yr-1 (Lidberg et al., 2010). As an example, in the parameters of the 14-parameter Helmert transformation from ITRF2005 to ITRF2000, the velocity of the Earth origin is -1.8 mm yr-1 along the Z-axis, and the rate of scale is 0.08 ppb yr-1 (ITRF, 2011). For a site at latitude 47°N, centrally located within the MRNF GPS network, this would imply a difference in vertical velocity of -0.8 mm yr-1 and -1.2 mmyr-1 in the north direction. This example shows limitations in the reference frame realization and the consequitive possible shift. Accordingly, one may not trust the superior TRF to the sub-mm yr-1 level without taking the evolution rates between TRFs into account.
Altamimi, Z., Collilieux, X., Legrand, J., Garayt, B., & Boucher, C. (2007). ITRF2005: A new release of the International Terrestrial Reference Frame based on time series of station positions and Earth Orientation Parameters. Journal of Geophysical Research, 112(B9), B09401. American Geophysical Union. doi:10.1029/2007JB004949.
Blewitt, G., Argus, D. F., Bennett, R. A., Bock, Y., Calais, E., Craymer, M. R., Davis, J. L., et al. (2005). A stable North America reference frame (SNARF): First release. UNAVCO-IRIS Joint Workshop Proceedings: Stevenson. Washington, USA.
Henton, J. A., Raymer, M. R., Ferland, R., Dragert, H., Mazzotti, S., & Forbes, D. L. (2006). Crustal motion and deformation monitoring of the Canadian landmass. Geomatica, 60(2), 173-191. Canadian Institute of Geomatics.
ITRF. (2011). Transformation Parameters between ITRF2005 and ITRF2000. International Terrestrial Reference Frame. Retrieved September 16, 2011, from http://itrf.ensg.ign.fr/ITRF_solutions/2005/tp_05-00.php.
Lidberg, M., Johansson, J. M., Scherneck, H. G., & Milne?, G. A. (2010). Recent results based on continuous GPS observations of the GIA process in Fennoscandia from BIFROST. Journal of Geodynamics, 50(1), 8-18. Elsevier. doi:10.1016/j.jog.2009.11.010