What about randomizing some values on every load to make existing, and additional values (as described in the article), result in new (and therefore useless) fingerprints on every load?
Your assumption seems to be that these changes won’t be recognized. But they are very easy to recognize, e.g. by grouping the values by IP address. Same IP address but constantly changing random values? Yes, that’s a very reliable fingerprint.