The scanning speed for NMap scans has seen some attention recently.
While the new version has a sticky problem at very slow speeds (I can't
find the link into the mailing list but it involves SYN scans and Sneaky
speed), there is also a paper
which discusses optimization of scanning times.