When using SerialGC or ParallelGC(and non-numa) interpreter never refills TLAB. After the first TLAB allocations we end up with bump the pointer on eden which can lead to performance issue with contented CAS operations. Both JFR memory leak and google low overhead profiler cannot profile with the inconsistent handling of TLAB refills in Interpreter.