Prove abrupt recovery with live jobs in released and managed containers

Assistant: codex
Assistant-Model: gpt-6-astra
Assistant-Session: 01a13590-a16f-7f60-be60-fd71627bc99d
This commit is contained in:
tegwick 2026-10-13 22:32:10 +02:00
parent 5b8e798419
commit 2cb2b654e0
25 changed files with 1954 additions and 4 deletions

View file

@ -165,3 +165,20 @@ cancellation. Wrong release and wrong execution location each fail before job
submission. Retained [verification](../docs/evidence/2026-10-13-consumer-verifier.json).
The command does not establish authentication, restart workloads or close T01;
pair it with native identity and permission receipts plus remaining recovery tests.
## Abrupt recovery passed — 2026-10-13 22:29 Europe/Berlin
User authorized the remaining recovery verification. Added and executed isolated
released-container and approved managed-pilot crash verifiers. A real screenshot
was present in an active job and another job was queued before SIGKILL. Managed
container exited137 and restarted on the same pod/PVC/image; both runs became
interrupted, partial files disappeared, completed PNG hashes matched and a fresh
14-step/four-PNG trial passed. Dedicated test connections were stopped. See
[executed recovery evidence](../docs/evidence/2026-10-13-crash-recovery.md).
The earlier abrupt-interruption gap is resolved. T01 remains wait for actual
human/agent identity/transport receipts and cross-version rollback; no second
release was invented solely to claim rollback. T03 remains dependent on managed
consumer acceptance. No new identity-owner implementation return was available
at this execution; CUST-WP-0080-T05 retains coordination.