Mongrel failures but nothing in logs

Hi All,

We are having a few strange problems in production.

The Rails 2.3/ Ruby 1.8.6 application is hosted using Apache/ mod_proxy/ mongrel_cluster (managed by monit) running on CentOS. This connects to a PostgreSQL instance running on a Windows 2003 server (it's behind a load balancer but the PostgreSQL port is redirected to one machine.

Suddenly, without reason or upgrade, the Rails app is returning 500 and nothing is in the Rails log - not in the mongrel logs, not in the production log. We seem to be able to safely connect to a static image file - so, it should not be the Apache, we think.

We checked a few things:
* mongrel has the rights to folders like tmp and log
* the disk can be written to

Since nothing is in the logs, we are baffled - what could we check?