No good reason for why it hasn't changed, except that the underlying single-threaded allocator (Doug Lea's malloc) is really good. The multithreaded version, which is a "per-thread" (not really) wrapper around the Lea allocator has serious problems that have been known for a long time. I wrote a paper outlining and demonstrating the problems in 2000. The inertia is difficult to explain.

Paper here: http://people.cs.umass.edu/~emery/pubs/berger-asplos2000.pdf

Hoard replacement allocator here: http://www.hoard.org/

(Note that Hoard's design has evolved considerably since 2000; the Mac OS X allocator is based on an older version.)