The criticisms I've heard typically reference pause times in the tens or hundreds of milliseconds. Agreed that different domains and applications have different

The criticisms I've heard typically reference pause times in the tens or hundreds of milliseconds. Agreed that different domains and applications have different requirements for pause times. I would be very interested to see histograms of pause times for different GCs, but I'm pretty sure the specific results would vary a lot depending on the specific corpus of applications under benchmark. If your 99%-ile pause time is tens of microseconds, is that good enough for video games? Audio kernels?