Invert HRRR failure logic: only retry transient network/server errors

Instead of listing every permanent failure string, only retry on known
transient errors (5xx, 429, network exceptions). Everything else —
GRIB decode errors, 404s, range request failures, index out of range —
is cancelled immediately.

lib/microwaveprop/workers/hrrr_fetch_worker.ex

@@ -56,28 +56,29 @@ defmodule Microwaveprop.Workers.HrrrFetchWorker do
           :ok
 
         {:error, reason} ->
-          if permanent_failure?(reason) do
+          if transient_failure?(reason) do
+            Logger.error("HRRR transient error for #{lat},#{lon} @ #{valid_time_str}: #{inspect(reason)}")
+            {:error, reason}
+          else
             Logger.warning("HRRR permanent failure for #{lat},#{lon} @ #{valid_time_str}: #{inspect(reason)}")
             {:cancel, reason}
-          else
-            Logger.error("HRRR fetch error for #{lat},#{lon} @ #{valid_time_str}: #{inspect(reason)}")
-            {:error, reason}
           end
       end
     end
   end
 
-  defp permanent_failure?("HRRR idx HTTP 404"), do: true
+  # Only retry on transient network/server errors — everything else is permanent
-  defp permanent_failure?("HRRR grib HTTP 404"), do: true
+  defp transient_failure?(%{__exception__: true}), do: true
-  defp permanent_failure?("HRRR range request not supported"), do: true
+  defp transient_failure?("HRRR idx HTTP " <> status), do: server_error?(status)
-  defp permanent_failure?(:outside_grid), do: true
+  defp transient_failure?("HRRR grib HTTP " <> status), do: server_error?(status)
-  defp permanent_failure?("malformed section"), do: true
+  defp transient_failure?(_), do: false
-  defp permanent_failure?("missing sections: " <> _), do: true
+
-  defp permanent_failure?("not a GRIB2 message"), do: true
+  defp server_error?(status) do
-  defp permanent_failure?("unsupported GRIB edition " <> _), do: true
+    case Integer.parse(status) do
-  defp permanent_failure?("GRIB2 extraction failed: " <> _), do: true
+      {code, _} when code in [429, 500, 502, 503, 504] -> true
-  defp permanent_failure?("GRIB2 complex packing decode failed: " <> _), do: true
+      _ -> false
-  defp permanent_failure?(_), do: false
+    end
+  end
 
   defp maybe_add_derived_params(attrs, nil), do: attrs