It appears that the initial assumption of writing to the data file, and then the index being a safe way to avoid data corruption may not hold in many circumstances (http://danluu.com/file-consistency/).
Chronos needs to be able to detect a corrupt index, and avoid corrupting it further. Additionally it should probably support a recovery mechanism for re-writing the datastore to remove corrupted entries.
It appears that the initial assumption of writing to the data file, and then the index being a safe way to avoid data corruption may not hold in many circumstances (http://danluu.com/file-consistency/).
Chronos needs to be able to detect a corrupt index, and avoid corrupting it further. Additionally it should probably support a recovery mechanism for re-writing the datastore to remove corrupted entries.