During partial network partitions on the downstream cluster, nodes on different sides of the partition might connect simultaneously to the upstream. That causes upstream exchanges to be recreated with different names (A/B) and links trying to add bindings to exchanges that don't exist anymore. That causes crashes like the one below and restarts the link.
Apparently, the best way to recover the link is to allow it to restart, but the crash can be captured and replaced by a more informative log message.
During partial network partitions on the downstream cluster, nodes on different sides of the partition might connect simultaneously to the upstream. That causes upstream exchanges to be recreated with different names (A/B) and links trying to add bindings to exchanges that don't exist anymore. That causes crashes like the one below and restarts the link.
Apparently, the best way to recover the link is to allow it to restart, but the crash can be captured and replaced by a more informative log message.