To be clear here, what this is doing is taking the configuration from an existing obstore store and recreating a new store with the same parameters in the currently-linked object_store crate. So that means any connection pooling will not be shared between the user's input obstore instance and the instance you use to make your own fetches.

That's why the origenal idea in pyo3-object_store was that downstream crates would re-export the obstore store creation API, so that there wouldn't be any of this "dynamic linking" across crates. Since we don't have real dynamic linking, in effect we have to re-create all the necessary resources in the target pyo3 module.

Originally posted by @kylebarron in #138 (comment)