Biased locking was removed from OpenJDK HotSpot JVM some time ago. This move had its reasons, but it resulted in a number of significant performance degradations in existing code. Thus, when non-contended locking optimization can be added, it still benefits virtual machines. Parallel stop-the-world garbage collection is one of the earliest GC variants in HotSpot, and yet it is the best choice in terms of high throughput. SubstrateVM which powers GraalVM native image supports synchronization monitors and offers serial stop-the-world GC as a basic collector. Due to the increased popularity of pre-compilation and its support in modern frameworks, it is remarkable how playing classical optimization pieces helps to improve the performance of such code. BellSoft recently put thin locking and parallel stop-the-world GC implementations for public review in GraalVM. It is interesting that within the GraalVM project both were implemented in Java. We’ll explore implementation details, benchmark results, and application benefits.