We had a problem with a customer that got a 1 minute timeout when restarting their DJango webapp, claiming that Apache did get stuck. After a test, in the Apache log files we found:
[Thu Oct 11 09:35:18 2012] [error] (111)Connection refused: proxy: HTTP: attempt to connect to 192.168.0.10:48880 (192.168.0.10) failed [Thu Oct 11 09:35:18 2012] [error] ap_proxy_connect_backend disabling worker for (192.168.0.10) [Thu Oct 11 09:35:19 2012] [error] proxy: HTTP: disabled connection for (192.168.0.10) [...]
The problem is that the Apache proxy module, recognising that the service was unavailable, stopped redirecting requests to it for one minute.
The problem was solved adding a retry=0 parameter to the ProxyPass directive:
ProxyPass / http://192.168.0.10:48880/ retry=0 timeout=5 ProxyPassReverse / http://192.168.0.10:48880/