TSP on a complete graph. Would be easy for any normal planner, but hard for 
graphplan because it doesn't realize you can't visit n nodes in fewer
than n steps. [even though it does know that it can only do one action in
each time step and each action only satsifies up to one of the
"visited" goals, this info isn't used in any useful way.] The "L" lowerbounding
option tells graphplan to explicitly use this kind of reasoning. Specifically,
if you have a "clique" of t goals [maximal clique found greedily] such that
no two in the clique can be solved in the same step, then reason that
you will need at least t steps.  This option make it run LOTS faster.

(A B CONNECTED)
(A C CONNECTED)
(A D CONNECTED)
(A E CONNECTED)
(A F CONNECTED)
(A G CONNECTED)
(A H CONNECTED)
(A I CONNECTED)

(B A CONNECTED)
(B C CONNECTED)
(B D CONNECTED)
(B E CONNECTED)
(B F CONNECTED)
(B G CONNECTED)
(B H CONNECTED)
(B I CONNECTED)

(C A CONNECTED)
(C B CONNECTED)
(C D CONNECTED)
(C E CONNECTED)
(C F CONNECTED)
(C G CONNECTED)
(C H CONNECTED)
(C I CONNECTED)

(D A CONNECTED)
(D B CONNECTED)
(D C CONNECTED)
(D E CONNECTED)
(D F CONNECTED)
(D G CONNECTED)
(D H CONNECTED)
(D I CONNECTED)

(E A CONNECTED)
(E B CONNECTED)
(E C CONNECTED)
(E D CONNECTED)
(E F CONNECTED)
(E G CONNECTED)
(E H CONNECTED)
(E I CONNECTED)

(F A CONNECTED)
(F B CONNECTED)
(F C CONNECTED)
(F D CONNECTED)
(F E CONNECTED)
(F G CONNECTED)
(F H CONNECTED)
(F I CONNECTED)

(G A CONNECTED)
(G B CONNECTED)
(G C CONNECTED)
(G D CONNECTED)
(G E CONNECTED)
(G F CONNECTED)
(G H CONNECTED)
(G I CONNECTED)

(H A CONNECTED)
(H B CONNECTED)
(H C CONNECTED)
(H D CONNECTED)
(H E CONNECTED)
(H F CONNECTED)
(H G CONNECTED)
(H I CONNECTED)

(I A CONNECTED)
(I B CONNECTED)
(I C CONNECTED)
(I D CONNECTED)
(I E CONNECTED)
(I F CONNECTED)
(I G CONNECTED)
(I H CONNECTED)

(preconds (at truck A))

(effects (at truck A) (visited A) (visited B) (visited C) (visited D) 
(visited E) (visited F) (visited G) (visited H) (visited I))