I think that the title says it all. The `pd.merge` function can be compute intensive and can benefit (I think) from parallel computing. It does not appear to currently release the GIL. I can easily push my CPU to 100% but no higher when performing parallel joins.