[racket] dumping or forking racket processes

Thinking about long-running servers... Has anyone looked into 
core-dumping a Racket process image so that it can be loaded as a new 
process quickly (a la Emacs)?  Or looked into forking a pool of Racket 
processes (a la Apache) to use process isolation to be resilient in case 
of problems like memory leaks?

Background... I've recently converted a large Web application from CGI 
and SCGI, and it's worked quite well in interactive testing and 
automated load testing so far.  However, this got me thinking about 
backup plans, in case we find in later load testing, say, a case with 
long lists that the CGC (which we're stuck with for now) is too 
conservative about.

-- 
http://www.neilvandyke.org/