There are 3 common reasons why a Pod may be terminated on Autopilot: node upgrades, a cluster scale-down, and a node repair. PDBs and graceful termination periods modify the disruption to pods when these events happen, and maintenance windows and exclusions control when upgrade events can occur.
