Fleet HTTP reachability and recovery evidence

Code correction

The shared Fleet health resolver previously accepted aggregate healthy or degraded status as evidence of reachability when the HTTP status was missing. It now treats a fresh check without a valid successful HTTP status as down. Regression tests cover aggregate statuses and missing/malformed persisted payloads. This correction does not restore an unavailable instance or implement automatic task closure. Existing process fallback for absent/disabled/stale checks is unchanged; that fallback must not be used as recovery evidence.

The same server resolver supplies web and agent Fleet results. Inspection of the shared iOS/macOS source found no native Fleet health surface; there is no client layout or localization change.

Recovery evidence

A completed task, running connector, or machine heartbeat does not prove external availability. Before claiming recovery, confirm the intended instance lifecycle, origin readiness, public DNS, tunnel mapping, valid TLS, and the expected login or application response from outside the origin network. Collect three consecutive successful checks five minutes apart and verify that scheduled Hub checks remain green. Keep customer identifiers, incident timelines, and operator evidence in the restricted incident record rather than documentation distributed to every instance.