Shenandoah: the garbage collector that could (part 2)

Attention! To fully understand everything Aleksey will tell us in this talk, we strongly encourage you to watch the first part of "Shenandoah: the garbage collector that could": https://youtu.be/CnRtbtis79U

Once you deal with major GC phases and make them concurrent, the GC hiccups are then caused by the shorter, but still important (sometimes stop-the-world) activities in-between the concurrent phases. The issues include fast GC roots scans, interaction with GC-aware language features (for example, weak references), dealing with safepoint machinery pitfalls, heap management and interaction with operating system, etc. This talk dives deeper into actual problems faced by the low-latency collector like Shenandoah, what can JVM developers do about it, and what cautious application writers may consider when designing their low-latency Java applications.