Disk usage discrepancy?

Yes, all nodes have had a higher load than before (see Updates on Test Data) and more slow disk subsystems are now an issue.
And there are several solutions on the node’s side (optimizing a filesystem, adding more RAM or using an SSD as a cache layer, offload some loads to the SSD like databases, etc.), and also some improvements in the code (like a badger cache: Badger cache: are we ready?).
And of course, if some use the one disk for several nodes, now they know exactly why it was forbidden by ToS from the start. I do not say, that you do so, however, please, do not try.