Data scrubbing for internal topics

[mmedenjak]

In some severe cases ("operator bug where node 0 was restarted with an empty data volume and and empty seed_servers list, causing it to start a new log"), the log can be different between nodes even if the offsets are the same. The issue then is that the cluster is in such a state where raft doesn't see any problems, and we have to manually compare and copy raft logs between nodes. It would be good if we had some kind of sanity-check to see that the internal topics are consistent all-over.

See this thread.