Disk usage discrepancy?

usually mean the timeout, so disk is unable to response in time. Since it’s NTFS, you need to perform a defragmentation and enable it if it was disabled for this disk (it’s enabled by default).
If that wouldn’t help, then you may disable the lazy mode.

it will use more IOPS, but should finish without interruption and should update databases at the end.
However, you need to have all 5 filewalkers successfully finished their work for every trusted satellite.
All data from untrusted should be deleted though: How To Forget Untrusted Satellites