The setup to reproduce this bug is similar to the one for issue2605, but
note the additional "echo V > f; darcs revert -a":

darcs init R1
cd R1
echo A > f
darcs record -lam A1 f
echo B > f
darcs record -lam B f
# we now have R1=A1:>B
cd ..

darcs init R2
cd R2
echo A > f
darcs record -lam A2 f
echo V > f
darcs revert -a
darcs pull ../R1 -a
darcs obliterate -a -p A1
# we now have R2=A2:>B
cd ..

and now we do:

> cd R2 && darcs unrevert -a
Backing up
/tmp/duplicates-unrevert/Darcs2/PatienceDiff/duplicates-unrevert/R2/f(.~0~)
We have conflicts in the following files:
./f

This has nothing to do with issue2605. The conflict can be created more 
easily:

darcs init R
cd R
echo A > f
darcs record -lam A f
echo R > f
darcs revert -a
echo B > f
darcs record -lam B f
darcs unrevert -a
cd ..

Perhaps this is not a bug but a feature?

Okay, so what happens here is not that the unrevert bundle contains 
conflicts. Instead, we record a patch that conflicts with what the 
user reverted. Then, when we unrevert, we get a conflict. I guess this 
is not completely unexpected.