feat: qualify complete K1 replay on worker

README.md

@@ -143,20 +143,21 @@ The first passive replay derivative is now implemented as
 `missioncore.k1-local-surface/v1`. It binds the immutable `RAVNOVES00` pack,
 estimates a rolling local surface from map points plus compatible pose, and
 publishes height, slope, roughness, confidence and conservative observed
-surface/occupied/unknown evidence. All `526/526` available samples produced a
-diagnostic result; no free-space, command, navigation or safety authority is
-inferred. Leave-current-frame-out qualification produced `525` next-frame
-samples with `0.038543 m` p50 median residual and `12` strict temporal jumps;
-the p95 residual tail remains too large for planner use. Read-only replay
-triage narrows this to `37` attention frames in `21` episodes and four
-high-priority frames. The selected scene, clickable complete-recording timeline
-and source-frame review queue are visible in **Парк → Диагностика LiDAR**.
-The selected frame can now switch to a prior-plane residual view: it preserves
-the previous-TTL plane and overlays the current lower-cell evidence as
-in-band, above-plane and below-plane observations. On source frame `1254`,
-`36/98` evaluated cells are above the `0.16 m` band and none are below it,
-localizing the heavy tail without naming an object or granting safety
-authority.
+surface/occupied/unknown evidence. The complete pack contains `4,489` timeline
+frames, `3,928` available/valid LiDAR frames and `9,207,270` points; there were
+no stale-pose, insufficient-surface or fit-failed frames. Prior-only
+next-frame qualification produced `3,927` samples with `0.04065 m` p50 median
+residual and `205` strict temporal jumps. The p95 residual tail remains too
+large for planner use. The current complete derivative is the first item in
+**Парк → Диагностика LiDAR**.
+
+The earlier content-bound `526`-frame review slice remains available as
+historical laboratory evidence. Its deterministic triage contains `37`
+attention frames in `21` episodes and four high-priority frames. The selected
+frame can switch to the prior-plane residual view; source frame `1254` has
+`36/98` evaluated cells above the `0.16 m` band and none below it. This
+localizes a positive residual structure without naming an object or granting
+safety authority.
 
 The same estimator now runs behind
 `missioncore.k1-local-surface-shadow-runtime/v1`, a capacity-two
@@ -165,20 +166,22 @@ latest-wins queue with a bounded diagnostic result ring. LAB E27 passed a
 were consumed, maximum queue depth was `1/2`, no frame was replaced, p95
 processing latency was `14.777 ms`, and state, point classes, step candidates
 and scalar results matched the immutable replay derivative exactly. This is a
-recorded-source-paced execution gate only; physical K1 worker binding,
+recorded-source-paced execution gate only; physical K1 acquisition,
 free-space, commands, navigation and safety authority remain unavailable.
 
-The physical shadow seam is now implemented but not yet physically qualified.
-The authenticated external worker can bind decoded map-frame `lio_pcl` and
-`lio_pose` in either arrival order through
-`missioncore.k1-local-surface-pose-binder/v1`, then feed the accepted estimator
+The authenticated external-worker seam is now implemented and replay-qualified.
+It binds decoded map-frame `lio_pcl` and `lio_pose` in either arrival order
+through
+`missioncore.k1-local-surface-pose-binder/v1`, then feeds the accepted estimator
 without blocking camera cadence. Point input and estimator work are
-capacity-two latest-wins queues; pose history and result history are bounded.
-The pinned `e28_physical_k1_local_surface_profile.json` rejects the run unless
-at least 100 point frames bind and exact accounting, pose misses, queue drops,
-p95 result age, profile identity and shadow-only authority all pass. No
-physical acceptance result is claimed until that profile is executed against
-an attached K1 on the NVIDIA worker.
+bounded separately: the authenticated ingress has LiDAR depth eight, while the
+freshness-first estimator remains capacity two. LAB E28 replayed the complete
+recording at 1× through the actual NVIDIA worker: all `4,570` LiDAR and `4,598`
+pose events arrived with zero gaps, all point frames bound, and `4,206`
+results were persisted at `9.1945 Hz`. Explicitly superseded work was
+`364/4,570` (`7.965%`), p95 result age was `89.530 ms`, and all acceptance
+checks passed. Physical K1 acquisition remains a separate commissioning test;
+it is not required for replay algorithm development.
 
 The complete RELLIS-3D v1.1 release is now admitted there and its full
 `2,413`-frame validation split is available in **Полигон → Датасеты**. The