The dashboard usage metric reflects ZFS-level allocated blocks, not the logical size of live files, so a gap after large deletes is expected. Volumes are mounted with continuous discard, which reclaims some space over time, but the storage layer allocates in 16 KB blocks while ext4 frees in 4 KB blocks, so partially-freed blocks can never be returned and the metric settles permanently above what df/du report. The fstrim failure is also expected since containers lack the required capability. We don't perform manual host-side reclamation, as any reclaim would stop at that same block-size floor. The only way to fully reset a volume's footprint is to migrate the data into a fresh volume and delete the old one. In dollar terms the excess is roughly 6 GB at $0.15/GB-month, so about $0.90/month across both volumes.