Caesar III two-plot natural immigration route, 2026-11-17

Evidence: scratch/runs/20261117-caesar3-two-plots-natural-immigration-save/.

The retained preceding run was reviewed offline before launch. Its unnecessary third-plot MOVE entered the host queue at 227.6 ms, was purged at 502.6 ms, and preceded the next renderer poll at 594.2 ms. It had no native receipt; the trace does not establish a product cause. That original remains closed.

One newly authorized original attempted the changed ordinary route: title, New Game, name C, Assignment 1, two connected road tiles and two housing plots, then no more construction inputs and passive observation of immigration. The first road-placement press became terminal UNKNOWN, so the two plots and the passive observation phase were never reached. There was no replay, restart, third-plot MOVE, Save or native Quit.

The run reused the exact c993 base, 1e585 adapter/loader, d112 host imports and module 9099bb3f694d5ec6dac2886b70c1920fa33f39ad9258b1e662d1ebe05dc2da0a. All retained source hashes matched. All 114 media files were checked before the native clock and remained unchanged after shutdown. Existing native controls, readiness and release proofs were reused without broad qualification or compilation. The local caller regression passed with zero dispatches, including existing-review intent/hover, one fresh review and unique operation names. The runtime exercised bound hover successfully without duplicate review.

Personally reviewed whole scenes show Assignment 1 Aventine and the initial city at Dn 2000, population 0, January 500 BC. The road tool and green first road footprint were visible. Twelve finite actions completed with native USER receipts and independent release; this count includes hover and menu inputs, not twelve constructions. No completed road construction or city growth is proved in this run.

The thirteenth action, place-roadtile1, pressed at 398,482 using a scene 6158 ms old, within the unchanged 15-second game freshness guard. Host queue entries 25 (WM_NCHITTEST) and 26 (WM_LBUTTONDOWN) were present at 234.9 ms and remained until purge at 501.4 ms. The next observed renderer poll was 597.4 ms after intent. The receipt was CANCELLED, consumed false, native message null. Independent UP completed at 1486.7 ms. The 500/1750/3000 ms bounds were unchanged. No product defect or latency repair is claimed.

The permanent UNKNOWN latch closed the endpoint to all later requests. The independent release preceded that latch; the request log contains zero requests after it. Host cleanup promptly stopped the native process. Native, outer, persistent owner and independent watchdog waits have both EOFs; the runtime scope is empty, known owned process IDs are absent, and the listener is gone. Descriptors were closed. Host cleanup is not native Quit. No box was created: ID and expiry null, state NOT_CREATED, stop receipt NOT_APPLICABLE.

No production code changed; no locked build or new canonical-build claim was needed. Evidence retains exact Git pins, relevant uncommitted diff, owned run-input patches, selected named whole screenshots, request/input audits and hash readback. Foreign work was preserved. The concrete blocker is the immutable native-input confirmation boundary on the sole closed original. Occupied homes, immigration, population growth, native Save and native Quit remain unproved. Further original execution needs renewed authorization.