We recently discovered a memory issue that can cause BehaviorSpace to execute runs very slowly and using more processors than there are runs.
A BehaviorSpace experiment using larger-than-usual model runs executed very slow, with the number of processors in use varying from very few to more than expected. The problem was fixed by allocating more memory to NetLogo, by editing its NetLogo.cfg file. Memory allocation is discussed in the NetLogo FAQ: search for the item “How big can my model be?”. (If available on your machine, you can allocate many gigabytes of RAM to NetLogo.)