This can reuse the existing infrastructure to profile launches, it just needs to trigger an extra start/stop of the Spark profiler. Spark's default background profiler should also be suppressed when this option is enabled.

It would also be useful to be able to profile server joining.

Added in 5.23.1.