Just had a check - the fellow Xenotars (105/2.5, 105/2.
are
very difficult to tell apart on 3-D subjects, also regarding bokeh. It looks like the rear blur increases faster with the 2.8, owing to more pronounced spherical aberration when leaving the focus plave. The only noticeable effect (whe pixel peeping shots made with the Df) is that the green fringes produced by the 1.8 are more "concentrated", more narrow, with higher green colour saturation. This could mean more nervous bokeh.
Here is an example from which I drew the above conclusion. First the whole scene; the 1.8 shot (on the left) is slightly darker, by maybe 0.2 EV. Picture taken at f/2.8. Distance : about 2.2m. All illustrations are X/Y screenshots with X = left shot= 105/1.8; 100% crops except the general view.
Lower center: you'll see the green fringe difference on one cutlery element. Also, the letters on the bottle (already behind focus plane) have a more saturated black color on the left, while of course having fuzzy edges.
Right: the big letters on the milk bottle are slightly less fuzzy on the left pic.
Left: no significant difference can be seen on the wooden hedgehog. Near the top, there are two pale horizontal stripes on the left, merging to one on the right, so there is a tendency towards softer bokeh with the 105/2.5, but once again, I do not expect the differences (neither in bokeh, nor in sharpness) to play a significant role.
This is maybe why the Naerfoto website rates the 105/2.5 higher: similar results with smaller size and weight make the 105/2.5 more usable. The full aperture of the 105/1.8 (or even f/2.0) are not going to be useful very often due to high aberrations and noticeable bokeh outline - unlike for instance the full aperture of the Zeiss 135/2. One use case, also mentioned by Tristin, is night shots, where the full aperture is more usable than expected (low astigmatism, low bleeding), and the drawbacks (LoCA, low contrast, bokeh) a lesser issue.
