Comment by Aarvay

When choosing to apply parallelism, it should never be about the length of the collection but only the computational intensity of the function that's being applied. There's a considerable overhead in spawning processes and collecting the result from them