Currently the entire state of the cluster is stored in a single folder on flash memory on an SBC, with no backups or redundancy. If the board were to crash or the disk get corrupted, there would be no recovery mechanism.

I should improve the reliability by making the cluster less resistant to failure in this way.