fix: prevent stale capacity poisoning and deepen recent-only refill scans

Keep ordinary capacity races from lowering persistent per-account limits.
Deepen recent-only discovery within the existing age and quality gates.
Add offline regressions for recovery and reserved-slot refill.

modelhub_submmit_api/main.py

@@ -822,6 +822,17 @@ def build_adaptive_scan_stages(
                 "limit": max_models,
             }
         )
+    if not allow_older_than_recent_window:
+        # Reserved positions still need a deep search of eligible recent
+        # models. Widening the date window alone repeatedly scans the same
+        # short prefix (often already processed) on every poll cycle.
+        stages.append(
+            {
+                "name": "recent_window_full",
+                "updatedAfter": recent_cutoff,
+                "limit": max_models,
+            }
+        )
     deduped: list[dict[str, Any]] = []
     seen: set[tuple[str | None, int, str]] = set()
     for stage in stages:

modelhub_submmit_api/modelhub_client.py

@@ -1029,7 +1029,12 @@ class ModelHubClientPool:
                 self._capacity_probe_cooldown_until[index] = (
                     self._capacity_probe_cycle + self._capacity_probe_cooldown_cycles
                 )
-            else:
+            elif self._weekly_target_locked() == index:
+                # Only the scheduled calibration may lower a learned cap.
+                # Ordinary rejection can race with another agent filling this
+                # account; our cached count is not evidence of its capacity.
+                # Persisting it (even zero -> one) strands otherwise usable
+                # positions until the next calibration.
                 observed_capacity = max(1, self._effective_count_locked(index))
                 if observed_capacity < self._account_caps[index]:
                     self._account_caps[index] = observed_capacity