diff --git a/tests/__resources__/station_information.json b/app/__resources__/station_information.json
similarity index 100%
rename from tests/__resources__/station_information.json
rename to app/__resources__/station_information.json
diff --git a/tests/__resources__/station_location.csv b/app/__resources__/station_location.csv
similarity index 100%
rename from tests/__resources__/station_location.csv
rename to app/__resources__/station_location.csv
diff --git a/tests/__resources__/station_status.json b/app/__resources__/station_status.json
similarity index 100%
rename from tests/__resources__/station_status.json
rename to app/__resources__/station_status.json
diff --git a/app/config.py b/app/config.py
index 092765e..dca3b68 100644
--- a/app/config.py
+++ b/app/config.py
@@ -1,13 +1,9 @@
-from datetime import timedelta
 from pathlib import Path
 from typing import Annotated
 
 from pydantic import (
     AliasChoices,
-    DirectoryPath,
     Field,
-    FilePath,
-    HttpUrl,
     PostgresDsn,
     TypeAdapter,
     field_validator,
@@ -25,15 +21,17 @@ class Config(BaseSettings):
 
     model_config = SettingsConfigDict(frozen=True)
 
-    data_refresh_period: timedelta | None = timedelta(minutes=1)
+    check_mapping_lookups: bool = __debug__
+
+    data_refresh_period: float | None = None
+    """How often the station status data should be refreshed.
+
+    If ``None``, only local data will be used: no requests will be made to external APIs.
+    """
 
     # The $PORT environment variable is used by most PaaS to indicate the port the app server should bind to.
     port: int = 9090
 
-    reverse_geocoding_path: HttpUrl | FilePath = TypeAdapter(HttpUrl).validate_python(
-        "https://api-adresse.data.gouv.fr/reverse/csv/"
-    )
-
     user_content_storage: Annotated[
         PostgresDsn | Path | None,
         Field(
@@ -43,12 +41,6 @@ class Config(BaseSettings):
         ),
     ] = Path("content")
 
-    velib_data_base_path: HttpUrl | DirectoryPath = TypeAdapter(
-        HttpUrl
-    ).validate_python(
-        "https://velib-metropole-opendata.smovengo.cloud/opendata/Velib_Metropole"
-    )
-
     @field_validator("user_content_storage")
     @classmethod
     def normalize_postgres_dsn(cls, value: object) -> object:
diff --git a/app/load_tables.py b/app/load_tables.py
index 144079b..d82af9d 100644
--- a/app/load_tables.py
+++ b/app/load_tables.py
@@ -6,9 +6,10 @@
 import atoti as tt
 import httpx
 import pandas as pd
-from pydantic import HttpUrl
+from pydantic import DirectoryPath, FilePath, HttpUrl
 
 from .config import Config
+from .path import RESOURCES_DIRECTORY
 from .skeleton import SKELETON
 from .util import read_json, reverse_geocode
 
@@ -106,19 +107,35 @@ async def load_tables(
     config: Config,
     http_client: httpx.AsyncClient,
 ) -> None:
+    if config.data_refresh_period is None:
+        reverse_geocoding_path: HttpUrl | FilePath = (
+            RESOURCES_DIRECTORY / "station_location.csv"
+        )
+        velib_data_base_path: HttpUrl | DirectoryPath = RESOURCES_DIRECTORY
+    else:
+        reverse_geocoding_path = HttpUrl(
+            "https://api-adresse.data.gouv.fr/reverse/csv/"
+        )
+        velib_data_base_path = HttpUrl(
+            "https://velib-metropole-opendata.smovengo.cloud/opendata/Velib_Metropole"
+        )
+
     station_details_df, station_status_df = await asyncio.gather(
         read_station_details(
             http_client=http_client,
-            reverse_geocoding_path=config.reverse_geocoding_path,
-            velib_data_base_path=config.velib_data_base_path,
+            reverse_geocoding_path=reverse_geocoding_path,
+            velib_data_base_path=velib_data_base_path,
         ),
         read_station_status(
-            config.velib_data_base_path,
+            velib_data_base_path,
             http_client=http_client,
         ),
     )
 
-    with tt.mapping_lookup(check=__debug__), session.tables.data_transaction():
+    with (
+        tt.mapping_lookup(check=config.check_mapping_lookups),
+        session.tables.data_transaction(),
+    ):
         await asyncio.gather(
             session.tables[SKELETON.tables.STATION_DETAILS.key].load_async(
                 station_details_df
diff --git a/app/path.py b/app/path.py
new file mode 100644
index 0000000..c3ea661
--- /dev/null
+++ b/app/path.py
@@ -0,0 +1,4 @@
+from pathlib import Path
+
+APP_DIRECTORY = Path(__file__).parent
+RESOURCES_DIRECTORY = APP_DIRECTORY / "__resources__"
diff --git a/app/start_session.py b/app/start_session.py
index b9edb3b..7d58900 100644
--- a/app/start_session.py
+++ b/app/start_session.py
@@ -39,7 +39,7 @@ async def start_session(
     """Start the session, declare the data model and load the initial data."""
     session_config = get_session_config(config)
     with tt.Session.start(session_config) as session:
-        with tt.mapping_lookup(check=__debug__):
+        with tt.mapping_lookup(check=config.check_mapping_lookups):
             create_and_join_tables(session)
             create_cubes(session)
         await load_tables(session, config=config, http_client=http_client)
diff --git a/app/util/run_periodically.py b/app/util/run_periodically.py
index b8306a7..bcbd03a 100644
--- a/app/util/run_periodically.py
+++ b/app/util/run_periodically.py
@@ -1,7 +1,6 @@
 import asyncio
 from collections.abc import AsyncGenerator, Awaitable, Callable
 from contextlib import asynccontextmanager
-from datetime import timedelta
 
 
 @asynccontextmanager
@@ -9,15 +8,14 @@ async def run_periodically(
     callback: Callable[[], Awaitable[None]],
     /,
     *,
-    period: timedelta,
+    period: float,
 ) -> AsyncGenerator[None]:
-    period_in_seconds = period.total_seconds()
     stopped = asyncio.Event()
 
     async def loop() -> None:
         while not stopped.is_set():
             await callback()
-            await asyncio.sleep(period_in_seconds)
+            await asyncio.sleep(period)
 
     task = asyncio.create_task(loop())
 
diff --git a/pyproject.toml b/pyproject.toml
index 0885a6a..9e6182a 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -51,6 +51,7 @@ ignore = [
     "PLC0414", # Redundant imports are used for re-exporting (See https://peps.python.org/pep-0484/#stub-files).
     "S101",    # `assert` is useful when used correctly (https://realpython.com/python-assert-statement).
     "TCH",     # Pydantic needs to resolve type annotations at runtime.
+    "TID252",
     "TRY003",  # Not worth the annoyance.
 ]
 select = ["ALL"]
diff --git a/task-definition.json b/task-definition.json
index f7f2e00..0a01392 100644
--- a/task-definition.json
+++ b/task-definition.json
@@ -3,6 +3,12 @@
     "containerDefinitions": [
         {
             "name": "atoti-session",
+            "environment": [
+                {
+                    "name": "DATA_REFRESH_PERIOD",
+                    "value": "60"
+                }
+            ],
             "essential": true,
             "logConfiguration": {
                 "logDriver": "awslogs",
diff --git a/tests/conftest.py b/tests/conftest.py
index 8614fe6..b0e76f4 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -7,7 +7,6 @@
 from app import Config, start_app
 
 _TESTS_DIRECTORY = Path(__file__).parent
-_TESTS_RESOURCES_DIRECTORY = _TESTS_DIRECTORY / "__resources__"
 _PROJECT_DIRECTORY = _TESTS_DIRECTORY.parent
 
 
@@ -19,11 +18,8 @@ def project_name_fixture() -> str:
 @pytest.fixture(name="config", scope="session")
 def config_fixture() -> Config:
     return Config(
-        data_refresh_period=None,
-        reverse_geocoding_path=_TESTS_RESOURCES_DIRECTORY / "station_location.csv",
         port=0,
         user_content_storage=None,
-        velib_data_base_path=_TESTS_RESOURCES_DIRECTORY,
     )
 
 
diff --git a/tests/docker/_docker_container.py b/tests/docker/_docker_container.py
index a7eb2a8..1823411 100644
--- a/tests/docker/_docker_container.py
+++ b/tests/docker/_docker_container.py
@@ -1,4 +1,4 @@
-from collections.abc import Generator
+from collections.abc import Generator, Mapping
 from contextlib import contextmanager
 from uuid import uuid4
 
@@ -13,10 +13,12 @@ def docker_container(
     *,
     client: docker.DockerClient,
     container_name: str | None = None,
+    env: Mapping[str, str] | None = None,
 ) -> Generator[Container, None, None]:
     container = client.containers.run(
         image_name,
         detach=True,
+        environment=env,
         name=container_name or str(uuid4()),
         publish_all_ports=True,
     )
diff --git a/tests/docker/conftest.py b/tests/docker/conftest.py
index 9b0374e..50020fb 100644
--- a/tests/docker/conftest.py
+++ b/tests/docker/conftest.py
@@ -54,7 +54,14 @@ def session_inside_docker_container_fixture(
 ) -> Generator[tt.Session, None, None]:
     timeout = Timeout(timedelta(minutes=1))
 
-    with docker_container(docker_image_name, client=docker_client) as container:
+    with docker_container(
+        docker_image_name,
+        client=docker_client,
+        env={
+            # Test external APIs.
+            "DATA_REFRESH_PERIOD": "30"
+        },
+    ) as container:
         logs = container.logs(stream=True)
 
         while "Session listening on port" not in next(logs).decode():
diff --git a/tests/docker/test_docker.py b/tests/docker/test_docker.py
index 4ae5ff4..16ee3de 100644
--- a/tests/docker/test_docker.py
+++ b/tests/docker/test_docker.py
@@ -1,12 +1,20 @@
 import atoti as tt
 
 from app import SKELETON
-from app.util.skeleton import CONTRIBUTORS_COUNT
+
+from ..total_capacity import TOTAL_CAPACITY
 
 
 def test_session_inside_docker_container(
     session_inside_docker_container: tt.Session,
 ) -> None:
-    cube = session_inside_docker_container.cubes[SKELETON.cubes.STATION.key]
-    result_df = cube.query(cube.measures[CONTRIBUTORS_COUNT])
-    assert result_df[CONTRIBUTORS_COUNT][0] > 0
+    skeleton = SKELETON.cubes.STATION
+    cube = session_inside_docker_container.cubes[skeleton.key]
+    result_df = cube.query(cube.measures[skeleton.measures.CAPACITY.key])
+    total_capacity = result_df[skeleton.measures.CAPACITY.name][0]
+    assert total_capacity > 0, (
+        "There should be at least one station with one dock or more."
+    )
+    assert total_capacity != TOTAL_CAPACITY, (
+        "The data fetched from the external API should lead to a different capacity than the one of the local data since new stations have been created since the data was snapshotted."
+    )
diff --git a/tests/test_session.py b/tests/test_session.py
index 3544ebc..abee3a5 100644
--- a/tests/test_session.py
+++ b/tests/test_session.py
@@ -4,17 +4,17 @@
 from app import SKELETON
 from app.util.skeleton import CONTRIBUTORS_COUNT
 
+from .total_capacity import TOTAL_CAPACITY
+
 
 def test_total_capacity(session: tt.Session) -> None:
     skeleton = SKELETON.cubes.STATION
     cube = session.cubes[skeleton.key]
     result = cube.query(cube.measures[skeleton.measures.CAPACITY.key])
     expected_result = pd.DataFrame(
-        columns=[skeleton.measures.CAPACITY.name],
-        data=[
-            (45_850),
-        ],
-        dtype="Int32",
+        {
+            skeleton.measures.CAPACITY.name: pd.Series([TOTAL_CAPACITY], dtype="Int32"),
+        }
     )
     pd.testing.assert_frame_equal(result, expected_result)
 
diff --git a/tests/total_capacity.py b/tests/total_capacity.py
new file mode 100644
index 0000000..937b02f
--- /dev/null
+++ b/tests/total_capacity.py
@@ -0,0 +1,2 @@
+TOTAL_CAPACITY = 45_850
+"""The expected total capacity of the system when using local data."""