Suppose a scenario, where we have 2 threads doing the read operation,

it is possible that when the thread has errored out, in say read_exact method, now when it tries to notify the other thread, it is possible that the other thread is somewhere between while loop and claiming the cache mutex, so the notification is missed,

and since it errored out in read_exact, it has not set the state.reader to the right state and is None, which can potentially cause missed notification and an indefinite while loop.

we should also add retries in case we face the issue when doing read_exact, especially if the error is Interrupted or UnexpectedEOF.