There is a design issue with the fallback destination handling when the presence of a fallback receiver is known *before* the associated primary receiver. Please see issue https://issues.apache.org/jira/browse/DISPATCH-1786 and the attachment https://issues.apache.org/jira/secure/attachment/13028935/DISPATCH-1786-fallback-test-fail-analysis.txt
>From a design perspective why is the fallback deliberately ignored before the primary is known? Certainly forcing the primary to be known first is easier for life-cycle accounting but that leads to unwanted behavior later. @tross do you recall if this is actually correct and part of the design? If not then do you have any suggestions on the best strategy for fixing it?
