Because any score with higher initScore is always better than a score with lower initScore, deeper nodes (which have more initialized entities) are taken before higher nodes. As a result, Score First now behaves exactly as Depth First exploration. This is true for the small 4-NQueens example but at this moment I cannot prove if it's true in general. I consider this a bug as the change wasn't intended and the Score First exploration now performs worse than before (it makes more steps).

Please take a look at this commit if you want to see how the Score First exploration has changed.