The cross-site events should be retried in case of an exception. There is a case where the JGroups site-up events are delivered after the cross-site event which makes the events discarded. Retrying on exceptions improves/fixes that scenario.

The second improvement is to back off if the cache or cache manager isn't in the running state yet. In the current code, the site would be switched to offline but we can be optimistic about it and retry.