Most databases providing ACID properties assume that the hard-disk promises atomicity at the granularity of disk sectors (which are typically 512 bytes), and will always write disk sectors on the platter in the same order they were written to the OS.
Unfortunately many hard-disks on the market fail to meet these requirements.
The LSS goes to a lot of trouble to avoid data loss without making such strong assumptions on the hard-disk. It is permissible for disk sectors to only be partially written, and also for writing to the platter to be out of order. This is achievable because the LSS doesn't use Write Ahead Logging (WAL), and it employs 128 bit check point identity testing as well as a 32 bit CRC check on "flush units" that are read during the recovery scan.
The root block contains two independent copies of the root block data, written in strict alternation, and CRC checked.