I'm a little disapointed. PyPy has just recently been becoming a fast attractive alternative to CPython (or Cython compilation).
If they go the STM route and spend years hashing out the details only to be 2x as slow as CPython, then that cuts out a huge number of people who would like to practically use PyPy.
I think he might have meant 2x slower than without STM and not 2x slower than CPython. Also from everything I have read you would still have the option to not use STM at all and just use the GIL which would not come with the extra overhead.
If they go the STM route and spend years hashing out the details only to be 2x as slow as CPython, then that cuts out a huge number of people who would like to practically use PyPy.