Outcome recorded when the probe times out or rejects.

Defaults to false — the policy this helper was built for, where a backend that will not answer counts as unreachable. That is the right answer when the probe reads the thing whose health is being reported.

It is the WRONG answer when the probe reads a proxy for that thing and the proxy can fail independently: messaging-plugin's Service Bus probe reads the namespace's MANAGEMENT plane to report on its DATA plane, so a management endpoint that is firewalled, absent (the emulator ships no TLS listener for it) or merely slow used to report a live, publishing broker as down and drain the replica. Such a probe passes fallback: undefined and widens T, so "I could not determine this" stays distinguishable from "it is down".