[Alexbeav]

A recent psx_last_run_report.json from an earlier start can make the optional kit-start check pass even when the new game writes no report. Remove the report in both accepted locations immediately before each start. Stage2 must then receive fresh frames from that start.

The regression mocks the actual kit-start path: recent reports at the kit root and build-release must both be rejected, while a genuinely new report at either location is accepted. The unchanged base incorrectly accepts the old999-frame report; the fix rejects it. No actual program, GUI, audio or focus is started by these tests.

Validation on current main4b4f1d05f9e7634e4b44a52e15b29c92bf34527c, Windows/Python3.14:

• Exact recent-report regression fails before and passes after; fresh-report controls pass at both locations.
• Full repository unittest discovery:40tests, zero failures, two existing optional-package skips.
• Source diff/whitespace checked; existing line endings retained. No default engine, source selection, delivered player output or release artifact changes. A real setup/player-path run is not claimed.

Developed with AI assistance; validated as described (test evidence in PR body). AI writes the code and the PR, but I always test before I send something up. Happy to iterate on this process with your feedback.