From #763:

b04feb12 35358
splitcell 97917
b08dec23 116830
b07dec27_20091025 106891
fdemo 138379 # don't worry about this one
b07dec17 105507
gap-modeldb 43039
nrntraub 82894
lytton98 7399
lytton97 9889
wang1996 26997
prknj 17664
b09jan08 116838
lytton99 7400

For reference:

Download those models and fix the nocmodl issues. If you do one, the others will likely fall into line.

The prknj differences are so small (1e-8) that they can be ignored.
The gap-modeldb differences aren't so small but the lines on the graphs overlap so I think it is acceptable. And i will diagnose the differences when I get time.
I think the fdemo differences are due to unique random seed on every run.
splitcell has empty gout because of rand.c compile errors, e.g.

rand.c:452:4: note: previous definition of ‘z’ was here
   (z)[0] = LOW(l); (z)[1] = HIGH(l); }
    ^