I think one of them was that the CH query code did a full
dijkstra on the up and down graphs rather than stopping once it there is
some meeting point that is optimal.
From BMA:
That was just lazyness on my part. The improvement would result in a
constant factor speedup, more or less.
From David Turner:
From BMA: That was just lazyness on my part. The improvement would result in a constant factor speedup, more or less.