The serialized form of the j.n.URL includes the hashCode field.
Initially this field has value -1. Serialized form of the URL object will contain the hashCode field with the value -1.

When the hashCode() method is called in a first time, the hash code calculated and the result is set to the field. Serialized form of this URL object will contain the hashCode field with the value different from -1.

The following mini test shows that serialization of the one URL object will result in the different byte arrays:
------------------------------------------------------------------
    public static void main(String[] args) throws IOException, ClassNotFoundException {
        URL url = new URL("http://oracle.com");
        byte[] bytes1 = ser(url);
        url.hashCode();
        byte[] bytes2 = ser(url);
        System.out.println(Arrays.equals(bytes1, bytes2));
    }

    private static byte[] ser(URL url) throws IOException, ClassNotFoundException {
        try (ByteArrayOutputStream baos = new ByteArrayOutputStream();
            ObjectOutputStream oos = new ObjectOutputStream(baos)) {
            oos.writeObject(url);
            return baos.toByteArray();
        }
    }
------------------------------------------------------------------

Output will be:
------------------------------------------------------------------
false
------------------------------------------------------------------

Serialization/deserialization via RI won't cause and error, because once deserialized hashCode field will have value -1, it will be calculated when hashCode() method will be called.
However, it can cause errors during serialization/deserialization via different implementations.

Tested with JDK9 b53.
Failed JCK tests is under development.