📄️ List gate servers (deduped)
Returns one row per distinct gate server (deduped by `analytics_correlation_id`, latest event per server). Backed by the shared `events_dataset.external_events_json` firehose today; a dedicated table is a follow-up. Default lookback when neither `start_time` nor `end_time` is provided is **7 days** (matches gate's heartbeat cadence so a fresh open reliably lists all live servers). If only one bound is provided, the other is derived using the 7-day default. Same 30-day cap applies when both bounds are provided (`end_time - start_time > 30d` is rejected with HTTP 400).
📄️ List handled gate requests
Returns one row per served gate request. Each row is the fully-hydrated `GateRequestHandled_v1` event as stored, including per-plugin timings and per-round-trip HTTP metadata inline (no separate detail endpoint). Default lookback when neither `start_time` nor `end_time` is provided is **24 hours** — the endpoint is a high-volume activity log so the short default protects casual opens from scanning a full 30-day window. If only one bound is provided, the other is derived using the 24-hour default. Same 30-day cap applies when both bounds are provided.
📄️ List gate heartbeats (deduped)
Returns the newest heartbeat per distinct gate process — same deduplication shape as `/gate/server-starts` (partition by `analytics_correlation_id`, newest `publish_time` wins). The row's `sequence` counter lets a caller distinguish 'gate has been alive throughout' (monotonic per process) from 'gate restarted' (new correlation_id with sequence starting at 1). Default lookback when neither `start_time` nor `end_time` is provided is **7 days** — 'who is alive right now' framing. Same one-bound derivation and 30-day cap as the sibling endpoints.