Creating an index can take a long time. If it is interrupted (for example because the process was stopped or died), then it would be nice if after a restart reindexing would continue where it was stopped. I'm not sure how complicated this is.
There are some more potential problems that should be documented / tested:
- When creating a new index in a cluster, which instance creates the index?
- When creating multiple indexes at the same time, is the repository only scanned once (and not once per index)?
- The same when manually triggering a reindex using the "reindex" flag.