When running surefire with `forkCount=1` and `reuseForks=false`, the loaded `.class` files are cached across forks. Therefore, when one wants to analyze the behavior of specific tests (e.g., which `.class` files are loaded into the JVM by the test), results will be incomplete.
This has worked on older versions (I particularly tried `2.19.1`) and I am wondering whether this level of isolation was removed on purpose and if there is a way to manually return to the old default behavior of maximum isolation even for class loading.
- JDK 11 (Windows Server 2019 Datacenter)
- Surefire 3.0.0-M5