I agree with what you're saying. I think we're saying the same thing. I see that in the text you're quoting I only proposed a way to handle hotly written rows, and didn't address read rows being changed. This is also a problem.

My broader point is that with serializable you need to be aware of these bottlenecks in the database and you need to create a data model and access patterns such that permance hits are avoided as much as possible.