diff --git a/changelog.d/sr-1rgb.md b/changelog.d/sr-1rgb.md
new file mode 100644
index 0000000..777d2f6
--- /dev/null
+++ b/changelog.d/sr-1rgb.md
@@ -0,0 +1,5 @@
+### Fixed
+
+- `StatifierRouter.Addresses.reap/3` runs on SQLite: its stamp and delete
+  no longer use Postgres's `= ANY(...)`, which failed every reap of a
+  SQLite host on 0.8.0 and later, under an integer or a text key.
diff --git a/docs/adr/0002-addressing.md b/docs/adr/0002-addressing.md
index 5d4ee47..3681bd7 100644
--- a/docs/adr/0002-addressing.md
+++ b/docs/adr/0002-addressing.md
@@ -1433,3 +1433,24 @@ creates the table, and it is run outside the version walk.
 
 The 0.9.0 section of `CHANGELOG.md` names the key, the processor, the
 front, the two calls and the opt-in table.
+
+## Note (2026-09-30, sr-1rgb): the address sweep's writes run on SQLite
+
+A Note, not an amendment: it decides nothing and changes no decision,
+amendment or Note above it. The Note of 2026-09-27 accepting the sr-w58a
+Amendment says the private `stamp/3` and `delete/2` of
+`StatifierRouter.Addresses` bind the ids in `fragment("? = ANY(?)",
+a.id, ^ids)`. That form is Postgres's own; on SQLite the statement fails
+with "no such function: ANY", so a host on SQLite could not reap its
+address table from statifier_router 0.8.0 on.
+
+Both functions now name their rows in `fragment("? IN (?)", a.id,
+splice(^batch))`: an `IN` list with one bound parameter per id, which
+Postgres and SQLite both take, written in batches of at most 500 ids a
+statement (the private `in_batches/2` and `@ids_per_statement`), under
+SQLite's smallest limit on bound parameters. Each id is still bound as
+the table handed it over, so the Amendment's rule that the package
+never casts an id it binds holds as before, and a reap answers the same
+counts on Postgres as it did. The tests
+`StatifierRouter.SQLiteReapTest` reap on SQLite under the default
+integer key, under a text key of digits, and past one batch.
diff --git a/lib/statifier_router/addresses.ex b/lib/statifier_router/addresses.ex
index 2f14343..cfc2043 100644
--- a/lib/statifier_router/addresses.ex
+++ b/lib/statifier_router/addresses.ex
@@ -106,6 +106,7 @@ defmodule StatifierRouter.Addresses do
 
   @terminal [:completed, :failed, :cancelled]
   @default_limit 1_000
+  @ids_per_statement 500
 
   @typedoc """
   What one reap did: how many rows it stamped `terminal_seen_at` on, how
@@ -266,27 +267,40 @@ defmodule StatifierRouter.Addresses do
     DateTime.compare(DateTime.add(seen_at, horizon_ms, :millisecond), now) != :gt
   end
 
-  defp stamp(_config, [], _now), do: 0
-
+  # Both writes name their rows in an IN list of one bound parameter per
+  # id, which Postgres and SQLite both take, each id bound uncast as the
+  # rest of this module binds them. A list longer than
+  # @ids_per_statement is written in batches of that many, one statement
+  # each, which keeps every statement under SQLite's smallest limit on
+  # bound parameters (999) with room for the stamp's own time.
   defp stamp(config, ids, now) do
-    {count, _} =
+    in_batches(ids, fn batch ->
       from(a in Config.queryable(config, Address),
-        where: fragment("? = ANY(?)", a.id, ^ids) and is_nil(a.terminal_seen_at)
+        where: fragment("? IN (?)", a.id, splice(^batch)) and is_nil(a.terminal_seen_at)
       )
       |> config.repo.update_all(set: [terminal_seen_at: now])
-
-    count
+    end)
   end
 
-  defp delete(_config, []), do: 0
-
   defp delete(config, ids) do
-    {count, _} =
+    in_batches(ids, fn batch ->
       config.repo.delete_all(
-        from(a in Config.queryable(config, Address), where: fragment("? = ANY(?)", a.id, ^ids))
+        from(a in Config.queryable(config, Address),
+          where: fragment("? IN (?)", a.id, splice(^batch))
+        )
       )
+    end)
+  end
 
-    count
+  # The rows `write` counts over every batch of `ids`; no statement for
+  # no ids.
+  defp in_batches(ids, write) do
+    ids
+    |> Enum.chunk_every(@ids_per_statement)
+    |> Enum.reduce(0, fn batch, total ->
+      {count, _} = write.(batch)
+      total + count
+    end)
   end
 
   defp next(rows, limit) when length(rows) == limit, do: List.last(rows).id
diff --git a/test/statifier_router/sqlite_migrations_test.exs b/test/statifier_router/sqlite_migrations_test.exs
index 16c180f..82559a7 100644
--- a/test/statifier_router/sqlite_migrations_test.exs
+++ b/test/statifier_router/sqlite_migrations_test.exs
@@ -1,8 +1,9 @@
 defmodule StatifierRouter.SQLiteMigrationsTest do
   # The version walk on SQLite, through ecto_sqlite3, against a database
-  # file of each test's own: no Postgres, no SQL sandbox, nothing shared
-  # with the rest of the suite.
-  use ExUnit.Case, async: true
+  # file of each test's own: no Postgres, no SQL sandbox. The one repo
+  # process runs under the module's name, so every module that starts it
+  # is in the :sqlite_repo group, whose modules never run at once.
+  use ExUnit.Case, async: true, group: :sqlite_repo
 
   alias Ecto.Adapters.SQL
   alias Ecto.Migrator
diff --git a/test/statifier_router/sqlite_reap_test.exs b/test/statifier_router/sqlite_reap_test.exs
new file mode 100644
index 0000000..491a9a5
--- /dev/null
+++ b/test/statifier_router/sqlite_reap_test.exs
@@ -0,0 +1,234 @@
+defmodule StatifierRouter.SQLiteReapTest do
+  # StatifierRouter.Addresses.reap/3 on SQLite, through ecto_sqlite3,
+  # against a database file of each test's own, under the default key and
+  # under a text key built with the :primary_key option. The execution
+  # statuses come from StatifierRouter.StatusStore: the repo holds only the
+  # router's tables. In the :sqlite_repo group with
+  # StatifierRouter.SQLiteMigrationsTest: both start the one repo process.
+  use ExUnit.Case, async: true, group: :sqlite_repo
+
+  import Ecto.Query, only: [from: 2]
+
+  alias Ecto.Migrator
+  alias StatifierPersistence.Storage
+  alias StatifierRouter.Addresses
+  alias StatifierRouter.Binding
+  alias StatifierRouter.Config
+  alias StatifierRouter.Schema.Address
+  alias StatifierRouter.SQLiteRepo
+
+  # A host on SQLite under the repo's default key.
+  defmodule MigrateIntegerKey do
+    @moduledoc false
+    use Ecto.Migration
+
+    def up, do: StatifierRouter.Migrations.up()
+    def down, do: StatifierRouter.Migrations.down()
+  end
+
+  # A host on SQLite under a text key the database fills in. The ids are
+  # made of digits alone and lead with a zero, so a reap that cast one to
+  # an integer would lose the zero and match no row of the text column.
+  defmodule MigrateTextKey do
+    @moduledoc false
+    use Ecto.Migration
+
+    def up do
+      StatifierRouter.Migrations.up(
+        primary_key: [
+          type: :text,
+          default: fragment("(printf('0%011d', abs(random()) % 100000000000))")
+        ]
+      )
+    end
+
+    def down, do: StatifierRouter.Migrations.down()
+  end
+
+  @version 20_260_930_000_501
+
+  @now ~U[2026-09-30 12:00:00.000000Z]
+  @hour 3_600_000
+
+  setup do
+    database =
+      Path.join(
+        System.tmp_dir!(),
+        "statifier_router_sqlite_#{System.unique_integer([:positive])}.db"
+      )
+
+    start_supervised!({SQLiteRepo, database: database, pool_size: 1})
+
+    on_exit(fn ->
+      for suffix <- ["", "-wal", "-shm"], do: File.rm(database <> suffix)
+    end)
+
+    :ok
+  end
+
+  # Five parcels, one address row each: two delivered, one still on the
+  # van, one whose execution the store no longer holds, and one delivered
+  # and already stamped a day ago.
+  @statuses %{
+    "ex_pcl_6001" => :completed,
+    "ex_pcl_6002" => :completed,
+    "ex_pcl_6003" => :active,
+    "ex_pcl_6005" => :completed
+  }
+
+  defp config do
+    {:ok, config} =
+      Config.new(
+        repo: SQLiteRepo,
+        delivery: StatifierRouter.RecordingDelivery,
+        store: %Storage{adapter: StatifierRouter.StatusStore, opts: @statuses}
+      )
+
+    config
+  end
+
+  defp seed do
+    rows =
+      for n <- 6001..6005 do
+        %{
+          scope: "depot_north",
+          document: "parcel_delivery",
+          key: "pcl_#{n}",
+          execution_id: "ex_pcl_#{n}",
+          inserted_at: @now,
+          terminal_seen_at: if(n == 6005, do: DateTime.add(@now, -24 * @hour, :millisecond))
+        }
+      end
+
+    {5, _} = SQLiteRepo.insert_all(Address, rows)
+    :ok
+  end
+
+  # The reap's answer, or the message of what it raised, so a statement
+  # the adapter refuses fails the assertion that names it.
+  defp reap(config, bindings, opts) do
+    Addresses.reap(config, bindings, opts)
+  rescue
+    error in Exqlite.Error -> {:raised, Exception.message(error)}
+  end
+
+  # key => terminal_seen_at, for every address row left.
+  defp addresses(config) do
+    config
+    |> Config.queryable(Address)
+    |> then(&from(a in &1, select: {a.key, a.terminal_seen_at}))
+    |> SQLiteRepo.all()
+    |> Map.new()
+  end
+
+  defp ids(config) do
+    SQLiteRepo.all(from(a in Config.queryable(config, Address), select: a.id))
+  end
+
+  # The delivered-scan binding, whose horizon keeps a delivered parcel's
+  # row for an hour after a reap first sees it delivered.
+  defp bindings do
+    {:ok, binding} =
+      Binding.new(%{
+        id: "delivered_scans",
+        source: "depot_scans",
+        match: "event.kind == 'delivered'",
+        key: "event.parcel_id",
+        document: "parcel_delivery",
+        event: "delivered",
+        data: ["parcel_id"],
+        dedupe: %{by: :message_id, horizon_ms: @hour}
+      })
+
+    [binding]
+  end
+
+  # The same reap, twice: the first stamps the two parcels it is the first
+  # to see delivered and deletes the orphan and the row stamped a day ago;
+  # the second, an hour on, deletes the two it stamped.
+  defp sweep(config) do
+    assert reap(config, bindings(), now: @now) ==
+             {:ok, %{stamped: 2, deleted: 2, next: nil}}
+
+    assert addresses(config) == %{
+             "pcl_6001" => @now,
+             "pcl_6002" => @now,
+             "pcl_6003" => nil
+           }
+
+    an_hour_on = DateTime.add(@now, @hour, :millisecond)
+
+    assert reap(config, bindings(), now: an_hour_on) ==
+             {:ok, %{stamped: 0, deleted: 2, next: nil}}
+
+    assert addresses(config) == %{"pcl_6003" => nil}
+  end
+
+  describe "reap/3 on SQLite" do
+    # sabotage: put back stamp/3's and delete/2's `? = ANY(?)` fragment ->
+    # red, the first reap raised "no such function: ANY".
+    test "stamps and deletes under the default integer key" do
+      :ok = Migrator.up(SQLiteRepo, @version, MigrateIntegerKey, log: false)
+      config = config()
+      :ok = seed()
+
+      assert Enum.all?(ids(config), &is_integer/1)
+      sweep(config)
+    end
+
+    # sabotage: cast every id to an integer before it was bound in stamp/3
+    # and delete/2 -> red, the first reap stamped and deleted nothing.
+    test "stamps and deletes under a text key of digits, bound uncast" do
+      :ok = Migrator.up(SQLiteRepo, @version, MigrateTextKey, log: false)
+      config = config()
+      :ok = seed()
+
+      assert Enum.all?(ids(config), &(is_binary(&1) and &1 =~ ~r/^0\d{11}$/))
+      sweep(config)
+    end
+  end
+
+  describe "reap/3 on SQLite, past one statement's batch of ids" do
+    # sabotage: in_batches/2 answered the last batch's count alone -> red,
+    # the first reap answered stamped: 201 for 1201 rows stamped.
+    test "stamps and deletes every row of a reap that examines more ids than one statement binds" do
+      :ok = Migrator.up(SQLiteRepo, @version, MigrateIntegerKey, log: false)
+      numbers = 7001..8201
+      statuses = Map.new(numbers, &{"ex_pcl_#{&1}", :completed})
+
+      {:ok, config} =
+        Config.new(
+          repo: SQLiteRepo,
+          delivery: StatifierRouter.RecordingDelivery,
+          store: %Storage{adapter: StatifierRouter.StatusStore, opts: statuses}
+        )
+
+      for chunk <- Enum.chunk_every(numbers, 100) do
+        rows =
+          for n <- chunk do
+            %{
+              scope: "depot_north",
+              document: "parcel_delivery",
+              key: "pcl_#{n}",
+              execution_id: "ex_pcl_#{n}",
+              inserted_at: @now
+            }
+          end
+
+        SQLiteRepo.insert_all(Address, rows)
+      end
+
+      assert reap(config, bindings(), now: @now, limit: 2_000) ==
+               {:ok, %{stamped: 1201, deleted: 0, next: nil}}
+
+      assert addresses(config) |> Map.values() |> Enum.uniq() == [@now]
+
+      an_hour_on = DateTime.add(@now, @hour, :millisecond)
+
+      assert reap(config, bindings(), now: an_hour_on, limit: 2_000) ==
+               {:ok, %{stamped: 0, deleted: 1201, next: nil}}
+
+      assert addresses(config) == %{}
+    end
+  end
+end
diff --git a/test/support/status_store.ex b/test/support/status_store.ex
new file mode 100644
index 0000000..028e129
--- /dev/null
+++ b/test/support/status_store.ex
@@ -0,0 +1,23 @@
+defmodule StatifierRouter.StatusStore do
+  @moduledoc """
+  A storage adapter that answers one read, `fetch_execution/2`, from a map
+  of execution id to status handed to it when the store is built:
+
+      %StatifierPersistence.Storage{adapter: __MODULE__, opts: %{"ex_1" => :completed}}
+
+  An id the map does not name answers `{:error, :execution_not_found}`.
+  It lets a test reap address rows in a repo that holds no execution
+  table, the SQLite repo of `StatifierRouter.SQLiteReapTest`. Test-only
+  support code, not part of the package's public API.
+  """
+
+  @doc "The execution's status as the map names it, or `:execution_not_found`."
+  @spec fetch_execution(%{String.t() => atom()}, String.t()) ::
+          {:ok, %{status: atom()}} | {:error, :execution_not_found}
+  def fetch_execution(statuses, execution_id) do
+    case Map.fetch(statuses, execution_id) do
+      {:ok, status} -> {:ok, %{status: status}}
+      :error -> {:error, :execution_not_found}
+    end
+  end
+end