Done.

Very interesting, thanks! Looks good to merge then, and then we should notify
the Google group as discussed earlier.

I ran an experiment for configurations using pairs of heuristics to evaluate the effects of omitting the 
second heuristic:

http://ai.cs.unibas.ch/_tmp_files/seipp/issue714-v1-configs-comparison.html

The total coverage numbers show that it is beneficial to consider both heuristics for all pairs of 
heuristics for eager and lazy search. This is probably not surprising since both the old and new version 
always evaluate both heuristics for preferred operators anyway.

Here are the results for the affected portfolios:

http://ai.cs.unibas.ch/_tmp_files/seipp/issue714-v1-portfolios-issue714-base-issue714-v1-compare.html

Together with other configurations the effects of actually considering both heuristics become smaller. 
The only exception is FDSS-1 on barman where using both heuristics solves 20 additional tasks. 

Our continuous integration tests are passing, indicating that all "standard" configs and aliases conform 
to the new configuration requirements.

Pull request looks good to me, but please do an experiment with all
configurations to make sure we caught all changes, that the new configurations
work as extended, and to see how the "fixed" portfolios differ in behaviour from
the old ones. (Conceivably, the "correct" configurations might perform worse in
which case we might want to change the portfolios.)

Due to the way the interaction between the parser, portfolios and iterated
searches work, I suggest to run all configurations based on portfolios for the
full 30 minutes. I'd be happiest if we also tested all other aliases, but here I
think I short timeout (e.g. 1 minute) is enough.

Once this is merged, please write an email to the Google group to let people
know about the changed option syntax.

I have made a pull request at 
https://bitbucket.org/jendrikseipp/downward/pull-requests/68
and changed the affected configurations in the wiki.

Changing the configurations in our code, I noticed that some configurations part 
of seq_sat_fdss_1 probably never used the second heuristic for goal-distance 
estimates: "lazy_greedy(hff,hcg,preferred=[hff,hcg],...)".

Could you have a look at the pull request, please?

Since nobody raised any objections, I will probably start working on this after 
issue719 is merged.

It may take some getting used to, but I think getting rid of the "allow omitting
[] for singleton lists" rule could be a good thing.

When a user wants to run lazy_greedy with ff() and cea() he might forget the 
square brackets:

--search "lazy_greedy(ff(), cea())"

The parser happily accepts this and cea() is used to calculate preferred 
operators. We believe that one way to prevent such subtle usage errors is to
always require square brackets for list arguments, i.e., 
"lazy_greedy([ff()], [cea()])".

(Note that issue507 discusses an orthogonal change that would also help with this 
kind of usage error: requiring keywords for all arguments.)

What are your thoughts about requiring square brackets for all lists?