diff --git a/alembic/versions/0105_worker_lane_one_cap.py b/alembic/versions/0105_worker_lane_one_cap.py
index 01b4813..d36458b 100644
--- a/alembic/versions/0105_worker_lane_one_cap.py
+++ b/alembic/versions/0105_worker_lane_one_cap.py
@@ -70,11 +70,21 @@ _CAPS = (
 
 
 def upgrade() -> None:
-    # The constraint goes first: it names `slots`, so dropping the column out
-    # from under it fails on Postgres.
-    op.drop_constraint("ck_worker_lane_slots_within_cap", "worker_lane", type_="check")
+    # The constraints go first: they name `slots`, so dropping the column out
+    # from under them fails on Postgres.
+    #
+    # `op.f()` around each name, and it is load-bearing. Without it alembic
+    # runs the name through Base.metadata's naming convention, which prepends
+    # `ck_worker_lane_` to a string that already carries it — and the DROP
+    # goes looking for `ck_worker_lane_ck_worker_lane_slots_within_cap`, which
+    # no database has. That is #3275 exactly, from the other direction:
+    # alembic 0088 had to RENAME four constraints created with the same
+    # doubling. Caught here by the integration lane, run 7365.
     op.drop_constraint(
-        "ck_worker_lane_slots_non_negative", "worker_lane", type_="check",
+        op.f("ck_worker_lane_slots_within_cap"), "worker_lane", type_="check",
+    )
+    op.drop_constraint(
+        op.f("ck_worker_lane_slots_non_negative"), "worker_lane", type_="check",
     )
     op.drop_column("worker_lane", "slots")
     op.drop_column("worker_lane", "enabled")
diff --git a/tests/test_worker_lanes.py b/tests/test_worker_lanes.py
index b3ba4fc..14be621 100644
--- a/tests/test_worker_lanes.py
+++ b/tests/test_worker_lanes.py
@@ -246,22 +246,16 @@ def test_an_unrecognised_queue_set_still_gets_a_true_label():
 # --- the model's invariant ---------------------------------------------------
 
 
-@pytest.mark.parametrize(
-    "slots,cap,ok",
-    [
-        (0, 0, True),
-        (1, 1, True),
-        (1, 4, True),
-        (5, 4, False),   # slots above its own cap
-        (-1, 1, False),  # negative slots
-        (1, -1, False),  # negative cap
-    ],
-)
-def test_worker_lane_check_constraints(slots, cap, ok):
-    """The constraints live in the database, not only in the service, because
-    a row violating `slots <= slots_cap` is not a rejected request — it is a
-    lane the reconcile (step 3) will drive UP to a number the operator
-    capped."""
+def test_the_row_carries_exactly_one_constraint_now():
+    """`slots >= 0` and `slots <= slots_cap` went with the `slots` column on
+    2026-09-23 — there is one number left, and the only thing that can be
+    wrong with it is being negative.
+
+    Asserted as the WHOLE set rather than as a membership check: a constraint
+    left behind naming a dropped column is not a harmless leftover, it is a
+    table the migration cannot have produced, and the model would then
+    describe a schema no database has.
+    """
     from backend.app.models import WorkerLane
 
     constraints = {
@@ -269,17 +263,13 @@ def test_worker_lane_check_constraints(slots, cap, ok):
         if hasattr(c, "sqltext")
     }
     # The names carry the convention's `ck_worker_lane_` prefix ALREADY — the
-    # model declares them bare and Base.metadata's naming_convention applies it.
-    # Asserting the prefixed form is what pins the thing that actually went
-    # wrong once: alembic 0088 had to rename four constraints that shipped as
-    # `ck_x_ck_x_name`, because the migration pre-prefixed a name the
-    # convention then prefixed again (#3275). A bare-name assertion here would
-    # pass just as happily against a doubled one.
-    assert constraints == {
-        "ck_worker_lane_slots_non_negative": "slots >= 0",
-        "ck_worker_lane_cap_non_negative": "slots_cap >= 0",
-        "ck_worker_lane_slots_within_cap": "slots <= slots_cap",
-    }
+    # model declares them bare and Base.metadata's naming_convention applies
+    # it. Asserting the prefixed form is what pins the thing that actually
+    # went wrong once: alembic 0088 had to rename four constraints that
+    # shipped as `ck_x_ck_x_name`, because the migration pre-prefixed a name
+    # the convention then prefixed again (#3275). A bare-name assertion here
+    # would pass just as happily against a doubled one.
+    assert constraints == {"ck_worker_lane_cap_non_negative": "slots_cap >= 0"}
     for name in constraints:
         assert not name.startswith("ck_worker_lane_ck_"), f"doubled prefix: {name}"