There's no universal taxonomy of "this error is retryable, this one non-recoverable"; it's context dependent. As a boring example, I might write something that

There's no universal taxonomy of "this error is retryable, this one non-recoverable"; it's context dependent. As a boring example, I might write something that detects when a resource gets hosted, e.g. goes from 404 -> 200. The best I imagine you can do is be able to easily group each error and handle them appropriately.