diff --git a/arrow-pyarrow-integration-testing/tests/test_sql.py b/arrow-pyarrow-integration-testing/tests/test_sql.py
index 92782b9ed473..e2e8d66c0f29 100644
--- a/arrow-pyarrow-integration-testing/tests/test_sql.py
+++ b/arrow-pyarrow-integration-testing/tests/test_sql.py
@@ -421,6 +421,14 @@ def iter_batches():
     with pytest.raises(ValueError, match="test error"):
         rust.reader_return_errors(reader)
 
+    # Due to a long-standing oversight, PyArrow allows binary values in schema
+    # metadata that are not valid UTF-8. This is not allowed in Rust, but we
+    # make sure we error and not panic here.
+    schema = schema.with_metadata({"key": b"\xff"})
+    reader = pa.RecordBatchReader.from_batches(schema, iter_batches())
+    with pytest.raises(ValueError, match="invalid utf-8"):
+        rust.round_trip_record_batch_reader(reader)
+
 def test_reject_other_classes():
     # Arbitrary type that is not a PyArrow type
     not_pyarrow = ["hello"]
diff --git a/arrow/src/ffi_stream.rs b/arrow/src/ffi_stream.rs
index 7005cadc623c..865a8d0e0a29 100644
--- a/arrow/src/ffi_stream.rs
+++ b/arrow/src/ffi_stream.rs
@@ -281,7 +281,7 @@ fn get_stream_schema(stream_ptr: *mut FFI_ArrowArrayStream) -> Result<SchemaRef>
     let ret_code = unsafe { (*stream_ptr).get_schema.unwrap()(stream_ptr, &mut schema) };
 
     if ret_code == 0 {
-        let schema = Schema::try_from(&schema).unwrap();
+        let schema = Schema::try_from(&schema)?;
         Ok(Arc::new(schema))
     } else {
         Err(ArrowError::CDataInterface(format!(