To be clear, you will pay more unless you have enough scale to already run more than {num_AZs_in_region} nodes for your workload.
If you want to be sure you get paid if you’re offline, your workloads have to be spread across all AZs in the region.
Otherwise your subset of AZs can go down and because it’s not the whole region you won’t get paid.
Actual availability numbers is often secondary to Big Corp customers. The benefit of the CYA/SLA agreement, is that when shit does hit the fan, as the manager you get to shrug and say "well, we bought the improved SLA, not my fault Amazon failed to deliver".