Instead of trying all server ids in the supplied list whenever an error or timeout was encountered this change only retries when certain errors such as noproc, nodedown etc is encountered.
If all errors result in a retry genuine API errors such as {error, not_member} in response to a ra:remove_member/2 interaction would be hidden in the aggregate error that is returned once all members have returned.
Instead of trying all server ids in the supplied list whenever an error or timeout was encountered this change only retries when certain errors such as noproc, nodedown etc is encountered.
If all errors result in a retry genuine API errors such as
{error, not_member}
in response to ara:remove_member/2
interaction would be hidden in the aggregate error that is returned once all members have returned.Related: https://github.com/rabbitmq/rabbitmq-server/discussions/11029