Otherwise it will always trigger - most large workloads will have some sort of observability tool enabled, such as a tracer, and recording data from startup will likely have long vm operations caused by redefinitions of classes, leading to information that is very likely not relevant to the user.