We need to fix crashes (or at least catch them) before writing the indexer as a journal client, as crashes prevent committing the new position of the kafka consumer
(unless we blanket ignore all crashes or classes of crashes; but I do not think we should do that)