I didn't tag this as a feature request because I think it's worth discussing first whether it would actually be a good idea.
I've noticed that in some cases (e.g. when the files involved in the calculation are relatively small) it could benefit from parallel processing. That's it, running the script multiple times in parallel instead of serially, which is how it currently works.
Obviously, this would only make sense under the right conditions (e.g. when the calculation doesn't depend on other files), and it would be entirely up to the user to enable it via a flag that specifies the maximum number of concurrent calculations.
Do you think this could improve the speed at which script columns are calculated globally speaking?