The problem is that if the scrub operation results in a write, and the power fails during the write, then data may be corrupted. So either the drive needs to have power loss protection or scrubbing has to be disabled before power is removed.
Or the drive must be able to recover from a failed scrub operation.
The drive needs to be able to recover from a failed write which could be triggered by a scrub operation, which short of a RAID type of organization will be impossible, and maybe even that will not be enough.
An interrupted write to a multilevel storage cell destroys any previous data but there is another failure mode which also applies to SLC. The state machine which controls the write operation can run amok if the supply voltage drops out of tolerance, and attempt to write to completely unrelated pages. This exact behavior was observed in drive tests. If those unrelated pages happen to be metadata like the Flash translation tables, then the drive will likely become inoperable.
This is completely separate from protecting data in transit which is stored in buffers. If only data in transit was at risk, then interrupted write operations would be no different than those on hard drives and modern filesystems handle this just fine by updating metadata as a separate operation.
I do not know why the state machine which controls destructive operations cannot be made proof against power loss however so far all drives with power loss protection rely on sufficient stored energy to complete any write, and presumably erase, operation. For a while SandForce advertised controllers that did not require a capacitor bank for power loss protection but testing showed that they still destructively failed.