We were previously adding it later on to fail early on startup error. This didn't handle well "warming up" errors. Now that it was reachable and listening to requests already, we can just use the real retry limit from the get-go.
We were previously adding it later on to fail early on startup error. This didn't handle well "warming up" errors. Now that it was reachable and listening to requests already, we can just use the real retry limit from the get-go.