[monsterlag]

The output-resolution integration test still expected the old 305 MiB pool for non-4K modes. Merged #1202 deliberately raised that default minimum to 512 MiB while preserving larger scaled pools and explicit overrides. Update that exact expectation and explain its source; all seven resolution/menu cases and their other assertions remain unchanged.

Validation:

• Windows LLVM-MinGW and Linux g++: targeted strict runner: 2 passed, 0 failed, 0 skipped on each platform, exercising the seven output-resolution cases and existing 12-case pool fixture.
• The original integration test fails on its first Windows case. The unchanged 12-case fixture rejects the exact pre-Prevent Castle Town guard-house crashes with a larger render-target pool #1202 sizing function on both platforms at the three below-floor cases; scale and override controls remain intact.

This corrects test coverage only. It changes no runtime allocation or performance behavior and makes no GPU/game validation claim.

Hosted CI on fff881d: all eight validation checks passed, including Windows and Linux source checks, Windows renderer and Builder, and Linux AppImage. The two release-publishing jobs were skipped.

[NotRexed]

This is in main now: the same 512 MiB expectation in tests/test_output_resolution.py landed with a1c4bcd (the merge of #841), so closing. Thanks!