By the way, let's not forget, that we have an additional network access for updating.

Let's assume we have a check that some object has not changed. We can go either to DB or cache. Going to database probably would be less efficient due to longer reading. But every time we update the object we have to go additionally to the cache and update a record.

So in some cases I suppose distributed cache is not worth it. Especially because it adds extra complexity. It shouldn't be used recklessly for every case.