How do `Use`, `Run`, and `Map` differ when building an ASP.NET Core middleware pipeline?

For production middleware I make continuation and response ownership explicit. Code after await next(context) can observe the final status and duration, but it may be too late to change headers if downstream code has started the response. A middleware that catches exceptions must sit before the components it protects. A branch can accidentally bypass authentication, authorization, rate limiting, or observability if it is composed outside their intended scope. I test representative matched, unmatched, rejected, and exceptional requests, including whether downstream middleware ran and whether the response was already committed. For complex conditional behavior I prefer clear endpoint metadata or a small, named branch over nested predicates scattered through startup. The goal is a pipeline whose control flow can be read top to bottom and whose short-circuits are intentional rather than side effects of forgetting to invoke next.