Some devices are noisy AER correctable error devices. While notification is needed to diagnose potential future problems, warning/error fatigue is a legitimate concern, with a considerable issue of storage of repeated errors that are correctable (e.g. Replay Timer Timeout, Bad Data Link Layer Protocol, etc). You might have found Brainiarc7's gist regarding this topic. We'll dive a little deeper into the how and why here.
Here is a great example in the following kernel error message: