# PENGO-E013 — Fixed ACT3 egg-interaction probes

Finite diagnostic, not a new score search. Four routes plus one600-frame
neutral control were preregistered from the unchanged E003 frame7306.
All121 historical pinned sources and the original ROM hash passed before
execution. First source v1 failed JSON tuple/list manifest equality before
any emulator process; that source/registration and zero-cost failure receipt
are retained. Corrected source v1.1 was separately registered before MAME.

| Probe | Result | Target interaction reached | First observed fault |
| --- | --- | --- | --- |
| egg entry3, LEFT | Death during travel | No | frame7342, death animation |
| egg entry4, LEFT | Death during travel | No | frame7342, death animation |
| egg entry7, UP | Sixth walking macro blocked | No | None in recorded route |
| egg entry7, LEFT | Sixth walking macro blocked | No | None in recorded route |
| Root neutral600frames | Death during neutral extension | Not applicable | frame7384, death animation |

Neither egg interaction nor an egg-elimination mechanic was tested by these
four routes: an unreachable-in-practice stance must not be described as a
failed push or break. No route qualified for the protocol's conditional
unchanged-input native confirmation, so none was started under E013. No
rerouting, extra push, score ranking or remaining-budget search followed.

The upward routes expose a controller limitation in the dense observations.
The sixth MU command starts at7383, player(152,64), expecting(152,48).
At7398 the observed position is(152,49), and at7399 it is(152,47), both with
matching sprite coordinates and controllable phase. It passes the expected
y48 without an observed aligned decision point. The legacy controller waits
for exact target equality, keeps UP asserted, reaches the upper wall(152,32)
at7414, and times out at7533. This is an observed exception to the assumed
cell transition; its mechanism and native-speed reproduction are not yet
verified. It does not establish a general two-pixel speed rule.

Cost:37,656 physical frame evaluations in five fresh accelerated processes,
4.547s measured batch wall time. The neutral control's eventual life/respawn
changes are retained in its timeline. These diagnostic deaths do not alter
E003's already frozen zero-observed-deaths prefix or incomplete result.
No new best trajectory, completed ACT3, diamond evidence or full-game record.

Next prerequisite: preregister a separate **E014 fixed replay** of the
unchanged upward-route tape to confirm the missed decision point at native
speed, including screenshots, tile/object observations and private RAM diff.
Only then evaluate an opt-in controller correction under another protocol.

Evidence: data/pengo/e013/fixed-egg-probes-v1-1/registration.json,
manifest.json, result.json, control.json and controller-inspection-v1.json;
private native traces remain on doppia-3090. Complete-machine identity UNKNOWN.