Pengo E016: hatching and direct egg removal differ
Five recipes reproduced twice from reset at native speed:10 processes,56,196 physical frames,240.797seconds. Every pair matches declared trace bytes and sampled rendered pixels/exposed devices. Complete-machine identity remains UNKNOWN. v1 was prepared but never executed; v1.1 corrects the death-phase censor before any emulator job, preserving v1 inputs and registration.
The initial automatic hatch removes coordinates(1,5),(9,5),(3,3) from the egg table at607, changes the egg header6→3, and creates three agreeing actor sprites at those coordinates. The total objective count stays6. The rendered sequence shows cracking ice, eggs and generated Sno-Bee. Table disappearance during hatch is a transformation, not an eliminated objective.
Two scoped ACT3 direct interactions behave differently:
| Target egg | Healthy root | Input | Direct acknowledgement | Matched neutral control |
|---|---|---|---|---|
| (11,7) | 6471, player(11,8) | UP+BUTTON6472–6473 | At6473: objectives8→7, eggs5→4 | Counter stays8 and header5 through6610 |
| (7,11) | 6850, player(7,10) | DOWN+BUTTON6851–6852 | At6852: objectives7→6, eggs4→3 | Counter stays7 and header4 through6989 |
Coordinates are zero-based. The action cases show destruction animation and a500-point glyph at the affected cell; no replacement Sno-Bee is generated there during the measured window. The three existing mobile actor generations persist. The objective/egg decrease happens before the display animation finishes. HUD/RAM score remains8030 or8560 over these fixed windows despite the500-point glyph; the score-award timing and value are not accepted as a scoring rule here. A points overlay must not be read as ordinary ice, an egg, or a new objective.
The second action has death_collision_delay at6925; its neutral control has the same phase at6921. Those deaths occur after the6473/6852 direct event windows and are retained, not erased to label the entire suffix safe. The first action/control suffix has no observed death. A verified objective-reduction primitive is not proof that waiting afterward is safe.
The fixed-source census additionally observes egg-coordinate movement: an ACT3 egg moves from(7,9) to(7,11) before its later direct elimination. Slot reuse and coordinate removal alone do not establish identity; finite native sliding confirmation remains separate. Later-ACT capacity and unobserved paths stay UNKNOWN.
Scoped skill admission: direct egg-block destruction with a verified adjacent stance and blocked push geometry can produce immediate official removal; hatch preserves objective identity as a generated Sno-Bee. Do not impose a mandatory EGG→SNO-BEE→CRUSH path. The runtime pixel-agent still receives pixels only; this is privileged planner/validation evidence.
E005 remains gated by the latency sample, opt-in controller and validated event reader. It targets6→0 without optimizing score or time. Historical search and Pleiads reproduction remain unchanged.
