fix: stop a transient error erasing the FAA release history

- narrow the fallback to FileNotFoundError and move the monotonic assert out of the try:
  a rate limit, parse error or corrupt download previously rebuilt three years of registry
  from a single day and republished it as the whole dataset
- authenticate release reads and walk back through releases, matching the Transport Canada
  reader; the FAA path was unauthenticated at 60 requests an hour on shared runner IPs
- verify downloaded asset size

This is what the matrix's required: true flag on faa claims to protect, so the flag was
advertising a guarantee the script could not keep.

Generated-by: Claude Opus 5 <noreply@anthropic.com>

src/create_daily_faa_release.py

@@ -37,13 +37,20 @@ from derive_from_faa_master_txt import convert_faa_master_txt_to_df, concat_faa_
 from get_latest_release import get_latest_aircraft_faa_csv_df
 df_new = convert_faa_master_txt_to_df(zip_path, date_str)
 
+# Only a genuine first run may rebuild from a single day. A rate limit, a parse error or a
+# non-monotonic download_date must stop the run: this file becomes tomorrow's base, so
+# silently republishing one day erases the accumulated history.
 try:
     df_base, start_date_str = get_latest_aircraft_faa_csv_df()
-    df_base = concat_faa_historical_df(df_base, df_new)
+except FileNotFoundError as e:
-    assert df_base['download_date'].is_monotonic_increasing, "download_date is not monotonic increasing"
+    print(f"No existing FAA release found, bootstrapping from today only: {e}")
-except Exception as e:
+    df_base = None
-    print(f"No existing FAA release found, using only new data: {e}")
-    df_base = df_new
     start_date_str = date_str
 
+if df_base is not None:
+    df_base = concat_faa_historical_df(df_base, df_new)
+    assert df_base['download_date'].is_monotonic_increasing, "download_date is not monotonic increasing"
+else:
+    df_base = df_new
+
 df_base.to_csv(OUT_ROOT / f"openairframes_faa_{start_date_str}_{date_str}.csv", index=False)

src/get_latest_release.py

@@ -139,15 +139,29 @@ def download_latest_aircraft_csv(
         Path to the downloaded file
     """
     output_dir = Path(output_dir)
-    assets = get_latest_release_assets(repo, github_token=github_token)
+    github_token = github_token or os.environ.get("GITHUB_TOKEN")
-    try:
+
-        asset = pick_asset(assets, name_regex=r"^openairframes_faa_.*\.csv$")
+    for release in get_releases(repo, github_token=github_token, per_page=30):
-    except FileNotFoundError:
+        assets = get_release_assets_from_release_data(release)
-        # Fallback to old naming pattern
+        try:
-        asset = pick_asset(assets, name_regex=r"^openairframes_\d{4}-\d{2}-\d{2}_.*\.csv$")
+            asset = pick_asset(assets, name_regex=r"^openairframes_faa_.*\.csv$")
-    saved_to = download_asset(asset, output_dir / asset.name, github_token=github_token)
+        except FileNotFoundError:
-    print(f"Downloaded: {asset.name} ({asset.size} bytes) -> {saved_to}")
+            try:
-    return saved_to
+                # Fallback to old naming pattern
+                asset = pick_asset(assets, name_regex=r"^openairframes_\d{4}-\d{2}-\d{2}_.*\.csv$")
+            except FileNotFoundError:
+                continue
+        saved_to = download_asset(asset, output_dir / asset.name, github_token=github_token)
+        if asset.size and saved_to.stat().st_size != asset.size:
+            raise RuntimeError(
+                f"{asset.name}: downloaded {saved_to.stat().st_size} bytes, expected {asset.size}"
+            )
+        print(f"Downloaded: {asset.name} ({asset.size} bytes) -> {saved_to}")
+        return saved_to
+
+    raise FileNotFoundError(
+        "No release in the last 30 releases has an asset matching 'openairframes_faa_.*\\.csv$'"
+    )
 
 def get_latest_aircraft_faa_csv_df():
     csv_path = download_latest_aircraft_csv()