In the diagram, I think it is refering to a multy_AZ deployment and a read replica in the same region and 2 cross-region replication instances? Am I correct. I have not heard about a Multi_AZ with two stanby instances. Or I might be missing something in the explicanation.
I’ve also never heard of more than a single "secondary" replica in a region. I’ll bounce this to content development for review. Hopefully someone with more RDS knowledge than I can’t comment here as well.
I understand the confusion in the diagram. In practice, Mutli-AZ has one primary instance and one standby instance. If the primary has a problem, the standby is promoted to a primary and the old primary is demoted to standby. Same thing occurs when OS upgrades or patches happen…they sort of swap back and forth using DNS.
If the whole AZ goes away, the standby is again promoted to Primary and a new Standby is provisioned in a separate surviving AZ…IF you have included a third AZ in your subnet group. I recommend creating a subnet group with at least 3 separate AZs used….even for single-AZ in case you want to convert them to multi-AZ later.
Conceptually, the diagram is correct in that you always have a Standby instance floating in another AZ that you have defined in your subnet group.
Practically speaking, you don’t really need to know how Multi-AZ manages it’s standby instances behind the scenes but just know that it provides HA in a quasi-active/active manner.
There is a really good in-depth article about the process here: https://aws.amazon.com/blogs/database/amazon-rds-under-the-hood-multi-az/