How does Cucumber-JVM handle parallel test execution with the JUnit 5 Platform Suite?

The JUnit 5 Platform Suite engine, the current recommended way to run Cucumber-JVM, integrates with JUnit Platform's own parallel execution support, letting scenarios run concurrently rather than strictly one after another.

# junit-platform.properties
cucumber.execution.parallel.enabled=true
cucumber.execution.parallel.config.strategy=fixed
cucumber.execution.parallel.config.fixed.parallelism=4

With parallel execution enabled, each pickle (compiled scenario) can run on a separate thread, which is where Dependency Injection's per-scenario instantiation actually becomes essential for correctness, not just organization: if step definition instances or their injected shared state were accidentally reused across concurrent scenarios, parallel runs would produce inconsistent, hard-to-reproduce failures from one scenario's state leaking into another's.

Because of this, code that deliberately bypasses per-scenario isolation, static/shared mutable fields, singletons holding scenario-specific state, is the most common source of flaky failures that only appear once parallel execution is turned on, even when the exact same suite passes reliably in sequential mode.