I guess the argument against would be that it adds complexity for a use-case which is probably infrequent. How often does the average HTTP client request the s

I guess the argument against would be that it adds complexity for a use-case which is probably infrequent. How often does the average HTTP client request the same resource multiple times concurrently? In what subset of those cases does it actually matter if requests are redundant? Why waste your user's clock-cycles on an edge case?