Merged, thanks.

Looks good to merge.

No objections from my side.

The results look fine to me.

http://ai.cs.unibas.ch/_tmp_files/pommeren/issue698-v1-issue698-base-issue698-v1-compare.html
http://ai.cs.unibas.ch/_tmp_files/pommeren/issue698-v1-issue698-base-issue698-v1-total_time-blind.png

Should I merge this?

Fine with it, too.

There are probably people (like Bjarne Stroustrup) who would say that we're
using lvalue references in too many method signatures, but I don't think we need
to start rethinking our policy regarding these just now. But perhaps something
for the future when we have more experience.

The code looks good to me.

Here is a pull request on bitbucket (Jendrik agreed to have a look):
https://bitbucket.org/FlorianPommerening/downward-issues/pull-requests/29

Experiments are in the queue.

In issue688 we noticed that we can probably change

for (list<int> ops : operators_for_val) {

to 

for (list<int> &ops : operators_for_val) {

in successor_generator.cc.