# E041 — Scoped post-hoc reader admission after E040

E040 passes12 native fresh reset replicas of six fixed variants (161,045 physical frames/17 launches,279.968seconds). In the neutral pair, the final ACT4 Sno-Bee enters17 at11601, stays at raw coordinates24,240 while departure sprite codes advance49→55, and retires at11721 as the official counter1→0. PU exposes a one-frame18 retirement at11724 before0. No push contact/score increment occurs at removal, no player death. All six variants reproduce departure and finish the act. The first UNKNOWN screenshot/object dump and before/after RAM diff are retained privately. The semantic label BOUNDARY_ESCAPE is supported by the rendered departure sequence and actor/counter trace, not the state number alone.

The same neutral continuation pays5,000 points at11913 with the rendered GAME TIME table showing19seconds. Score26,040→31,040 crosses the30k extra-life threshold. Lives byte3→132 (0x84); HUD reserves2→3, hence4total lives. Only witnessed0x84 with exact three reserve glyphs, valid HUD score≥30k and in-game255 is decoded as4. Other high values and mismatching HUD remain UNKNOWN. No general masking rule is claimed.

Before new scoring, freeze and regress the new opt-in composition on all88 E038 traces, two M01 final proofs, two M02 native ACT3→4 handoffs and all12 E040 native traces. Preserve all original source contracts, legacy UNKNOWNs and crush/egg/diamond events/credits. Each E040 trace must have exactly one escape removal, no crush credit at that frame, raw17 accounting1, safe native post-bonus endpoint4lives, and cloning identity. Existing M01 bonus UNKNOWN phases remain excluded from search. Test corrupted state17 placement (interior), actor count2, wrong previous state/no prefix, wrong sprite color, death latch, and corrupted0x84 HUD/score/high-byte. These are synthetic negative annotations, never game results.

No emulator starts or physical frames;8CPU workers,15minutes. PASS_SCOPED_ESCAPE_LIFE only if regression adds0 new faults relative to the unchanged reader, supported contexts gain exact accounting, targeted negatives stay UNKNOWN/uncredited and source hashes match. Otherwise NUL; infrastructure separately INCOMPLETE. This is explicitly post-hoc admission, not universal phase correctness or independent pixel-agent accuracy. Historical modules remain frozen. A separate registration and two NEW native reset replays are required to promote the31,040 candidate; no automatic score promotion from E040.