That out-nmapping-nmap article is pretty funny. Here's something else I find odd about their proposed traceroute program:
"... we can send all TTL packets in one microsecond, instead of waiting for each one to come back before sending the next."

Correct me if I'm wrong but doesn't that presuppose that you know every route from anywhere to anywhere? If I ping some box (anywhere), won't it most likely do this:
my box>default gw(dgw) >dgw >dgw >bgp (or MPLS)>bgp >routing lookup (rl) >rl >rl >mac lookup >target box

To send "all TTL packets in one microsecond" you would have to know all those addresses listed above beforehand.

So for a standalone, distributable program, it would 1)have to be huge to hold all routes (and probably slow) & 2)update its routes constantly. Anyone that has watched one of the realtime BGP looking glasses would see how often routes change, not to mention redundant connections and load balancing.