Comment by froh

I'd rather love to see confluent persistence in python, i.e. a git-like management of an object tree.

so when you create a new call stack ( generator, async sth, thread) you can create a twig/branch, and that is modified in-place, copy on write.

and you decide when and how to merge a data branch,there are support frameworks for this, even defaults but in general merging data is a deliberate operation. like with git.

locally, a python with this option looks and feels single threaded, no brain knots. sharing and merging intermediate results becomes a deliberate operation with synchronisation points that you can reason about.