these two algorithms seems fast and efficient for aligning (superpose) two proteins:
Acta Chrystallogr. Sect. A 61 (2005) 478
[doi: 10.1107/S0108767305015266] J. Comput. Chem. 31 (2010) 1561
[doi: 10.1002/jcc.21439]
This is apparently a fast and efficient method to obtain the optimal rotation matrix. Being so new (particularly the second), I don't think VMD has this method implemented, and I wonder if it could benefit from it, especially when a great number of structures have to be aligned (as with the RMSDTT plugin).