Enigma supervisor output: independent acceptance

Exact candidate426bc58ffaa742c35fb5848a54b358c9d9cc7f3a is independently accepted without production changes. Six relevant suites passed on this exact code, followed by a separate independent real-program control suite. review.json binds24 inputs to actual committed Gitblobs. Full transformed/configured supervisor sources and exact deployment-adaptation receipts are adjacent.

The complete supervisor accepts a1229723-byte640x480 PNG with exact RGBA only under explicit single-frame2097152/cumulative2097152/max1. Default131072/cumulative393216/max3 is unchanged. Kernel artifact writes stop at the selected131072/2097152 limit. Oversized PNG rejects before artifact; writer EIO preserves original identity and latches later writes.

Fast/slow/EOF combined stdout+stderr retains at most131072bytes and fails shell-AND dependent work. Real TERM-resistant child/grandchild are KILLed/reaped within1.136seconds. Independent controls also cover split-boundary fault detection, log shortwrite/write/flush/close EIO, cleanup EIO and plain childexit7. Removing finaldrain, retention, or nonzero propagation exposes rejecting realprogram witnesses. The deliberate240000-byte mutant is rejected, never productioncap acceptance.

The full-program fixture changes only deployment admission time, UID, memory floor805306368, and CPU60. Exact adaptation is independently reconstructed byte-for-byte; original host diskfloor and production transport/finalization are untouched. Existing strict receipt validation and rejecting pin/anchor/profile controls remain; old mock-tail oracles now provide an explicit output interface and strengthen child/guard failure propagation. They do not stand in for complete-program acceptance.

Reviewer sparse dependencies were initially incomplete: terminal.log retains ENOENT; dependency-recovery.json records exact committed recovery. Production and testoracles needed no change.

After all correctness tests passed, actual disk admission rejected free2809901056/projectedafter2700849152<2982150144. Dependent integration stopped. Later bookkeeping observed an actual688865280-byte recovery; recovery-admission.json records fresh PASS projected45511240/67108864 after3389714432. The first failure and prior blocked-closeout artifacts remain intact. Existing authorization permits continued conditional serialized integration after this changed condition; no unchanged denial is retried.

For exact-code replay, use the six existing test files listed in review.json under the cheap256MiB heap/CPU60/wall60/output131072 supervisor in run-check.py. Set ENIGMA_SUPERVISOR_TEST_EVIDENCE to a fresh absolute directory for fullsource emission. independent.py records the separate realprogram rejection controls and exact fixture reuse; its original ops path is retained in the sealed receipt. Compressed supervisor sources expand with Python gzip; full-result.json binds their hashes. Standing codec identities are checked by each relevant fixture.

Source acceptance establishes no authentic Enigma frame/menu/gameplay. No guest/native/build/box/acquisition/deployment occurred. No box assigned; coordinator owns lifecycle. Historical producer evidence remains historical, including its failed final run; this independent finaltree run supplies the missing acceptance. Shared-lock integration must compare all tested code/fixtures to the committed inputs, reconcile actual remote main and verify pushed refs.