You can
set up your applications so
that, when a runtime error occurs during application processing, the
recovery service stores the events (specifically, asynchronous requests)
for later submission instead of repeatedly generating failed events.
For example, if you are deploying an application that calls a web
service and that web service is unavailable, a failed event is generated
and then subsequent associated asynchronous requests are stored. Then,
when the service becomes available, you can forward stored asynchronous
requests for processing.