# Testing That Cancellation Releases a Worker and Its Memory

A simple, reliable approach is to **combine a lifecycle signal with a memory measurement** around a cancellable task, then assert both drop back to baseline.

## The basic recipe

1. **Record a baseline** (worker count + memory) before starting