Matt Whiting wrote:
99 out of 100 stops are routine, but it is the 1% that can get an
officer killed. Having the occupants remain in the car gives them
more control of the situation until they have it scoped out fully.

Your stats are way off. Less than 1 in 300,000 stops are likely to get an
officer killed.

The supporting math: 54 officers died nationwide in shootings in _all_
aspects of their work[1] in 2006, including shootings not related to
traffic stops. There were ~16.7 million traffic stops reported nationwide
in 2002[2]. Assuming the stats between 2002 and 2006 did not change
significantly, that yields a rate of less than one death every 300,000
stops. Ironically, according to reference [1]:

"[...] traffic-related incidents claimed the lives of more officers (73)
than shootings (54) or any other cause of death. Of the 73 officers who
died in traffic-related incidents, 47 were killed in automobile crashes, 15
were struck by vehicles, nine died in motorcycle crashes, and two died in
bicycle crashes."

Obligatory aviation content from that same reference:

"Other causes of deaths included job-related illnesses (18), aircraft
crashes (3), beating (1), stabbing (1) and terrorist attack (1). Five of
the officers killed during 2006 were women."

Aircraft killed more cops in 2006 than terrorists! (In the U.S. ....)

[1] http://www.nleomf.org/media/press/LODYearend06.htm
[2] http://www.ojp.usdoj.gov/bjs/pub/pdf/cdsp02.pdf