Part D — teardown, three layers (2026-10-21 09:38–09:45 UTC)
Machine (demo-hp, Tier 0) — DONE, by own IDs only (demo-hp.txt): VM 9320 rehearsal1020-box1 destroyed with its
disks; CT 9330 rehearsal1020-browser destroyed; ISO rehearsal1020-felhom-installer-1.29.0-pve9.2-1.iso removed; no
drill nftables table left (only the product's felhom_oob); nothing of the run on nvme-scratch. Memory available
15 320 MB (start: 15 612 MB, ../A3-capacity/before.txt). Untouched: Tester 1 (VM 341), 9201, 9202, 9401. The 9202
data-loss proofs were removed through the product during C3 (their 03-teardown.txt).
DooPlex: the three watchers stopped by their own PIDs; nothing else ran here but builds and git.
Hub — FOR THE OPERATOR: host Tester4-648b80 (the rebuilt box, now destroyed) is still on the Hosts page. Delete it there; until then it alarms as stale/down (expected mails). The customer Tester4 stays (the standing fresh-install customer). The old host Tester4-816e21 was deleted by the operator at ~09:11.
ep0 and the Storage Box — listed, nothing removed (read-only, ep0-readonly.txt):
- ep0 PBS datastore
felhom-offsite, namespaceTester4: one snapshotct/9201/2026-10-20T12:17:26Z(the dead box's install-time whole-box copy; 68 KB of index, chunks shared). WireGuard peer 10.77.0.6 (Tester4) still configured, with the stale 2028 handshake of R-1204. - Storage Box sub-account u629488-sub5,
/home/felhom-repo: the household's restic history — the hub's own listing at 05:10Z saw 44 snapshot files (newest 02:18Z); the clock tests then added snapshots dated 2026-10-25 and 2026-10-28. Not read directly (no credential in this run); not removed. It is the history R-1176 cannot reach.
Secrets: 24 names REHEARSAL_2026_10_20_* in sec (sec-names.txt, names only). Kept; the operator decides. Seven
are spent one-time values (three bind tokens, two pairing codes, two setup codes).