It's well known that speed differences on the road are a major cause of safety problems, so unless you are going to have segregated self-driving corridors, having them operate at a higher speed than other traffic is dangerous even if the speed would be safe for them if the other traffic was also self-driving at similar speeds and safe for the non-self-driving traffic if the other traffic was non-self-driving at the lower speed.
>It's well known that speed differences on the road are a major cause
the speed differences you are referring to are road design situations such as highways dumping traffic into downtown traffic. delta V isn't recognized as a severe problem by IIHS nor DOT. speeding very much is.
--What is the effect of speed differences?
In addition to absolute speeds, the speed differences between vehicles also have an effect on the crash rate. This effect is studied in two ways. The first type of studies are those that compare the crash rates between roads that have a large speed variance (large differences in vehicle speeds during a 24 hour period) and roads that have a small speed variance. These studies mostly conclude that roads with a large speed variance are less safe (Aarts & Van Schagen, 2006). The second type of studies are those that concentrate on the speed differences between the individual vehicles that were involved in a crash and all the other vehicles. The first studies of this type were conducted in the United States in the 1950s and 1960s, e.g. Solomon (1964). These studies always
found a U-curve: the slower or faster a car drives compared with most of the vehicles on that road, the more the risk of being involved in a crash increased. However, more recent studies, especially those carried out in Australia (e.g. Kloeden et al., 1997; 2001; 2002) that used more modern measuring instruments and used a more accurate research design, reached a different conclusion. They still indicate that vehicles that drive faster than average on that road have a higher crash rate; vehicles
that drive slower, however, were found not to have an increased risk (Figure 3)