From pythomnic3k

Retried calls in Pythomnic3k

Retried call is essentially a message sent from one cage to another, or from a cage to itself. Rather than being an arbitrary object, the message has similar semantics to an RPC call - it contains module name, method name and call arguments and results in the specified method being called. The difference is that RPC is synchronous and retried call is asynchronous:

The key difference between regular RPC call and retried RPC is that that the latter is stored to a persistent storage and retried until it succeeds, surviving cage restarts and power failures. The only way for a retried call to stop is for its target method to return successfully, without throwing an exception.