Pengo E020: reader fix alone insufficient

E020 is terminal: NUL, progress-timeout. With only the egg observation changed, the frozen E003 planner retained3lives, reduced ACT3’s official counter from8 to5 and then committed25decisions without another reduction. It stopped at frame7104 with9,490observed points,57committed decisions,3,522fresh MAME processes and23,215,441physical frame evaluations in1,994.485search seconds. The final trajectory has not yet undergone native-speed final verification. It is not a new verified record.

The corrected egg representation alone was insufficient for ACT3 completion in this frozen configuration. It produced more observed objective reduction than historical E003’s8→6, but this comparison does not identify an exclusive cause or demonstrate robustness on other mazes. The causal claim remains scoped to the controlled intervention. All sources, action order, heuristic, visits, original oracle_v2 controller,20workers,width4/depth4 and25-decision official-counter watchdog were held fixed. E018 and the new causal policy were excluded. No unused budget or retry was consumed after the registered stop.

The original preregistration stays immutable: protocol, public source manifest, full private manifest SHA256 68c23aafb42b9281febb1ec7d05ff92bffcf0510db7970bbebadc4397b5862d8. Terminal outcome metadata omits raw RAM/object state and the ROM.

Category: TRAJECTORY SEARCH / READER-ONLY COUNTERFACTUAL / PRIVILEGED. Complete-machine identity remains UNKNOWN. Normal stagnation is a NUL result; historical E003’s frozen label remains unchanged.9,790 remains the best verified planner trajectory, with9,390last alive and death from7299; it is not a full-game record.

The next separately preregistered configuration is E005: Objective Reduction Planner, using causal event credit and reachable push stances. It does not resume E020 or optimize score. The pixel-only decision-model experiment remains stopped on prerequisites.