Pythomnic error-handling strategy is influenced with the nature of its application - processing separate synchronous requests incoming from different interfaces.

Essentially, an exception unhandled in your application code propagates up to the interface which has initiated the request, where it is caught, wrapped up and delivered with the response using the interface-specific mechanism. This is easier to demonstrate with a web service, for example if an HTTP request fails, the client gets a "500 Internal server error" response with a chunk of text similar to this:

UnboundLocalError("local variable 'undefined' referenced before assignment") 
in incoming_request() (http_interface_receiver.py:27) 
<- takes_invocation_proxy() (typecheck.py:379)
<- __call__() (module_loader.py:764)
<- _handle_request() (http_interface.py:80)

With a message-oriented interface, such as JMS, the protocol doesn't support responses, therefore the interface simply writes the exception text to the log file and aborts the message processing by rolling back the receiving transaction.