Earlier on the Tor development community seemed to have been much more focused on preventing attackers who can see exit traffic (the exit node, destination sites) from linking exit traffic to a single entity, than they were focused on preventing attackers from tracing targets in relatively short amounts of time. This is somewhat evident in the fact that circuits used to rotate once every thirty seconds, which would have resulted in attackers being able to trace people roughly twenty times faster than they can today (particularly as there were not even entry guards initially). The reason they changed from thirty seconds to ten minutes was not even because of this fact, but rather because the volunteer nodes had trouble keeping up with the computational demands of doing so many cryptographic operations to create new circuits so quickly. I believe that recently they have been giving more thought to increasing untraceability, and they currently have plans to extend circuit rotation time past ten minutes, reduce the number of selected entry guards and add layered entry guards. I am quite glad that they are heading in this direction, as I believe that a large majority of those who use Tor would prefer the chance for very strong anonymity/privacy at the risk of having no anonymity/privacy versus having weaker anonymity with a stronger guarantee of some anonymity/privacy. This is particular true for people like vendors on SR, because like I said, even if your circuit is pwnt for two seconds you are just as fucked as if it is pwnt for two hours.