1. A new benchmark is added, nobody cares
  2. A newbie with 2 hours of Lisp experience notices that there's
     no implementation for the benchmark. He thinks that a crappy
     implementation is better than no implementation, and proceeds
     to write something embarrasingly slow and hideously ugly.
  3. People who mistakenly think that the Shootout has any validity
     bitch and moan about Lisp being slow on that benchmark.
  4. Someone who knows how to optimize Lisp code gets tired of the
     moaning, and either rewrites the whole program (if it was
     sufficiently hideous) or makes a few obvious but critical
     improvements (if the program was salvageable). A huge speedup
     results, the people from step 3 start whining about something else.
  5. The requirements for the benchmark are modified, and the optimized
     Lisp implementation gets deleted. There's no sign of it ever
     having existed.
  6. Return to step 2

The fannkuch benchmark is currently at step 3, having been at step 5
at least once. The old fannkuch program was about 10 times faster than
what is currently up on the shootout site.