Merged. I can't say I reviewed this thoroughly -- I don't understand that code
at all, and I think it should be one of our next refactoring targets. But at
least the patch looks plausible. :-)

Thanks, Erez!

No; will try to look into it today.

Malte - did you merge this?

Pathmax did not update the heuristic value which was passed to the open list, so 
it was never working right.
This problem was due to a strange interaction between LM-A* and pathmax.
This is fixed in my downward-fixes repository.

This happens with psr-small:p16-s29-n2-l5-f30.pddl, with the config
--search 
"astar(selmax([lmcut(),lmcount(lm_hm(m=1),admissible=true)]),mpd=true,pathmax=tr
ue)"

I am investigating