It turns out that PR #35687 causes a regression when non-existent resources are specified, leading to early failure. Since existing third-party code relies on late/conditional failure there (https://github.com/spring-projects/spring-framework/pull/35687#discussion_r2560620853), we should make this lenient and accept non-existent resources still, simply adding them at the end of the scripts list (since we cannot properly sort them).