Yeah, additional context from timestamps would only help make better decisions.
Certainly a 404 w/o a timestamp is less valuable than a 404 w/ a timestamp, and 204 w/ a timestamp < than 404 w/ a timestamp less valuable still. The more context we can bring in the better. But the more I think about those handoffs the more I think we need a way to make more than replica# requests. An explicit (per-policy) config var would be sufficient.
Yeah, additional context from timestamps would only help make better decisions.
Certainly a 404 w/o a timestamp is less valuable than a 404 w/ a timestamp, and 204 w/ a timestamp < than 404 w/ a timestamp less valuable still. The more context we can bring in the better. But the more I think about those handoffs the more I think we need a way to make more than replica# requests. An explicit (per-policy) config var would be sufficient.