Upstream Prematurely Closed Connection

[APseudoUtopia]

I recently upgraded PHP from 5.2.9 to 5.3.2. Ever since then, nginx
hasn't been working properly. BUT, I can tell that it is not php's
fault because another domain is working perfectly (using the same
nginx instance).

nginx version: nginx/0.7.65

admin.mysite.com works perfectly. I have some PHP scripts on that
subdomain, and they work perfectly.

However, mysite.com gives a 502 HTTP error. The error logs in nginx give:

2010/04/14 21:57:54 [error] 93347#0: *35 upstream prematurely closed
connection while reading response header from upstream, client:
96.x.x.x, server: mysite.com, request: "GET / HTTP/1.1", upstream:
"fastcgi://127.0.0.1:9000", host: "mysite.com"

nginx is serving other non-php files perfectly fine (for example,
images and javascript). But all php files give the 502 error.

The php-cgi processes are started using spawn-fcgi. I have 20 child
processes set to run. I have monitored them throughout many requests
and none of them seem to be dying - they all exist after the 502
errors.

Here's my entire nginx.conf. There are a couple servers: the admin
site, a couple ssl/www redirects, then the main site. PHP on the admin
site works fine, but PHP on the main site does not work as mentioned
above.
http://pastie.org/private/hwi3niwgyynmcpvlgkuuxg

Here's the fastcgi_params include:
http://pastie.org/private/edbmxdtqkazapisegq6ua

Does anyone know why this is happening?

Thank you.