We use it :P
We're running worldoflogs.com on top of mod_wsgi, using parts of the django framework. We've tried all other deployment methods, but none of them can match mod_wsgi's memory footprint and stability in multi process + threaded mode.
We serve around 600k GET and 300k POST requests per day on 1 Dell PE1950, using around 20% of CPU and 2G memory at peak traffic around midnight, 5 mbit in (data uploads), 5 out (pure html).
When we were still new to this a few months ago, the site crashed one day because it ran out of memory - there's a big, big memory leak somewhere in the version we ran, and apache child processes used more and more memory until they got killed by MaxRequestsPerChild. After setting that on 100 as a quick fix, we evaluated alternative hosting methods.
FCGI wasn't selected for two reasons - configuration complexity and no multi-threading support. Even after reading all the docs, it wasn't clear what exactly I had to do to get it running, so I tried mod_wsgi first. As for multi-threading, you just can't run a single python instance with 10 threads with fcgi, but as most time is spent waiting for data from external sources, it isn't a very efficient use of memory. That was also one of the mod_python issues, we run apache with mpm-prefork.
Mod_wsgi on the other hand was a breeze to setup. After adding the server config, we never really had to look at it again, it just did its job and didn't bother us. We did change the configuration once to accommodate more concurrent clients, but that's just changing a number and reloading the apache config.
