How does PA addressing work with multiple uplinks in IPv6 without NAT?

Question

Computer networks

In IPv4, everything was simple. On the node that performs address translation, when the main channel was broken, translation to the addresses of the backup one began. But what about IPv6? When switching the channel, the addresses of thousands of machines must change at once? Or it is necessary to register AS with PI on such a network? But then this all directly contradicts the goal of IPv6 - to reduce the amount of information in global routing systems.
And a related question: There is a TLA block in the address for very large carriers. As I understand it, Tier I. But Tier I do not distribute addresses to subscribers. Accordingly, it is not clear why the largest pieces of addressing are reserved for them.

Answer 1

nat is a crutch, in IPv6 you just need to have your own ASN and a few BGP uplinks

Answer 2

For small sites without their own ASN and PA block, you can really use NAT66 (only it is called NPTv6 and is a stateless solution).
On the other hand, hosts can have several IP addresses on the interface, which means that two addresses can be distributed via dhcp, one from one block, the other from another. The problem will be when one operator breaks. We will have to catch this break and revoke the issued addresses to hosts from the corresponding pool. It's long. The concept is here .
This was not the intent of IPv6. And NAT in IPv4 does not affect this in any way.