| Original baseline's zero-swap intent did not match effective unlimited swap. |
Candidate uses tested 52 GiB RAM / 1 MiB swap assertion; recreated baseline enforces zero swap. |
Candidate containment verified; original cause unproven. |
| APC cache capacity fell 6.78%. |
Kept context/C4 configuration; verified 309422-token C1 needle and 110760-token tool call. |
Bounded gate passed; full-context C4 remains unmeasured. |
| Original 64-word baseline failed 2/8 exact counts. |
Retained failed native artifact; froze strict 16-word protocol for both arms. |
Failed cell excluded from performance. |
| Eight-request unique scout regressed. |
Retained negative signal; frozen 32-request finalist met 5% throughput gate. |
Comparison complete, bounded variance caveat. |
| Original quality runs had two repetitions below minimum three. |
Reran both arms with three repetitions and typed control provenance. |
12/12 each passed. |
| Candidate unique finalist omitted native configuration fingerprint. |
Retained external artifact-hash and actual identity binding. |
Traceability retained; no native field invented. |
| Initial systemd inactive boolean was inaccurate. |
Final timestamps prove unchanged active(exited) wrapper predating campaign. |
Corrected explicitly; no unit mutation. |
| New report lacked one hardware-index entry. |
Added the campaign plan to the hardware audit. |
Documentation fix. |
| Test subprocess could not find bare python. |
Supplied the existing virtual environment on PATH. |
Environment correction; no product code change. |