Should botocore retry on HTTP 408?

[sstoops]

Reading the AWS Documentation, it lists 400 and 408 as transient errors that could potentially be retried. See this link under Retryable.

However the botocore documentation does not list these two status codes as retryable. Checking the source for TransientRetryableChecker verifies this fact.

I stumbled upon this 2012 discussion where a user named garnaat, presumably a boto dev, mentioned that at one point, AWS requested retrying on 408 be removed. They suggested if the AWS docs states 408 codes should be retried, then boto should be updated.

I haven't been able to find any further discussion on the subject. Should boto be updated to match the official docs?