This paper presents another analysis of the sizes and surface densities of early-type galaxies at z=1-2; here they use a sample of 32 spectroscopically-confirmed galaxies with a mean redshift of 1.45 from several different surveys. Masses and ages are determined via SED fitting to the photometry. As has been reported before, these galaxies lie well off the z=0 size-luminosity relation. By evolving the galaxies (assuming pure luminosity evolution) from their measured redshifts to the present, the authors find that some galaxies would actually evolve to the z=0 relation in this manner, while some would not. These galaxies appear to be "young" and "old" respectively, and the authors appear to claim a bimodal age distribution in ETGs at this redshift, with typical ages ~1 Gyr and 3.5 Gyr. Young ETGs follow the local size-mass relation, old ETGs do not.
The authors conclude that the young objects have more or less completed their evolution (except for luminosity evolution), while the old galaxies still need some process to increase their effective radii. Dry merging cannot do this because it would create too many high-mass galaxies, so some other process must be at work. Much of this rests on the assumption that the relative ages can be accurately determined through photometry, of course.