Why is it considered hard to maintain strong consistency in a distributed system?

Question

Why can\'t you just use strongly consistent reads for all your DB reads, with retries on 500 responses? According to CAP theorem increasing consistency should probably lower