Let me add: The key point of the GC per process is that when heaps are small, pause times are also small. And since GC is counted against a process reductions f

Let me add: The key point of the GC per process is that when heaps are small, pause times are also small. And since GC is counted against a process reductions frequently GCing processes get swapped out all the time and will not hurt the latency of other processes.