1) Rebalancer should not print 'object is outdated'. It should handle this error internally, without scaring users. part of #578
2) Don't print 'Finish bucket recovery step, 0 ... buckets are recovered'. It does not help, but clogs the logs.
3) Log own UUID when started, because no other way to understand who to whom sends buckets.
4) Try to log bucket_id when they are recovered. That would help to find what a bucket was duplicated and how.
5) Rebalancer and other built-in subsystems should gracefully handle an error, when peer instance of a connection wasn't configured yet. It leads to errors like "Procedure 'vshard.storage.rebalancer_request_state' is not defined". part of #578