A stop-the-world period in managed runtimes (JVM, Go, Node) where the application freezes while memory is reclaimed, the silent latency tax.
A garbage collection pause is the time during which a managed runtime (Java's JVM, Go's runtime, Node's V8, .NET CLR) freezes the application threads to reclaim unused memory. Modern collectors (G1, ZGC, Shenandoah, Go's concurrent GC) keep pauses under 10ms for most workloads, but heap-sized or allocation-heavy services can still see multi-second pauses that look like outages from the outside.
GC pauses are the most common cause of 'random' p99 latency spikes that don't correlate with anything else: no traffic spike, no downstream slowness, just the runtime stopping the world to reclaim memory. Tuning the GC (heap size, collector choice, allocation patterns) is one of the highest-leverage performance investments for JVM and Go services. Tracking GC-pause time alongside latency is what makes the connection visible.
See the part of the platform that handles garbage collection (gc) pause in production.