Catalogs don't like to get overwritten, because write_file will error if it notices there's a duplicate catalog. The goal is, instead of erroring for a named catalog, the old one must be versioned using its hash. Once the old catalog is moved, the new one can take its place with no error so long as the new file is different from the old file. An error should still occur for hashed catalogs, since those are content-addressable, and so there's no reason to do the same work twice.
Catalogs don't like to get overwritten, because write_file will error if it notices there's a duplicate catalog. The goal is, instead of erroring for a named catalog, the old one must be versioned using its hash. Once the old catalog is moved, the new one can take its place with no error so long as the new file is different from the old file. An error should still occur for hashed catalogs, since those are content-addressable, and so there's no reason to do the same work twice.