How do the different modern garbage collectors (G1, ZGC, Shenandoah) differ in their approach to minimizing GC pause times, and when would you choose one over the default?

G1 (Garbage-First, the default collector since Java 9) divides the heap into many small, equally-sized regions and prioritizes collecting the regions with the most garbage first, aiming to meet a configurable target pause time goal (-XX:MaxGCPauseMillis) while still handling both young and old generation collection, suitable for the vast majority of general-purpose applications with heaps up to tens of gigabytes; ZGC and Shenandoah are more specialized, low-latency collectors designed to keep pause times extremely short (typically sub-millisecond to a few milliseconds) even for very large heaps (hundreds of gigabytes or more) by performing the vast majority of GC work concurrently with application threads rather than requiring long application-pausing phases, at the cost of somewhat higher CPU overhead and memory footprint compared to G1 -- ZGC/Shenandoah are typically chosen specifically for latency-sensitive applications (like high-frequency trading or real-time systems) where even G1's already-good pause times aren't sufficient.

# Selecting a garbage collector via JVM flags
java -XX:+UseG1GC -XX:MaxGCPauseMillis=200 -jar app.jar        # default, general-purpose, tunable pause target
java -XX:+UseZGC -jar app.jar                                   # ultra-low latency, very large heaps
java -XX:+UseShenandoahGC -jar app.jar                           # similarly low-latency, different implementation approach