When a descriptor gets rate limited, its key is stored in the local cache as OVER_LIMIT, then the next time a new request with the same key is received, it is automatically rate limited. However for descriptors that have a dynamic limit override this can have unintended behavior. If the first request comes in with a dynamic limit of a low threshold and gets rate limited, the key gets stored in the local cache as OVER_LIMIT. Later, if a request comes with the same descriptor but a different dynamic limit of a higher threshold, it will automatically get rate limited even though the new threshold has not actually been reached.
Is there a way to stop this behavior for descriptors with a dynamic threshold without turning off the local cache completely?