In rare scenarios we have the bootstrap member becoming the leader just before we're terminating the machine.
The bootstrap member itself does not have the time to move its leadership somewhere else, likely because it's not terminated by a SIGTERM.

This causes unnecessary leader elections during the bootstrap and apiserver unavailability 

4.17 and 4.18, but might have existed before    

rarely    

    1.
    2.
    3.
    

no apiserver unavailability due to the bootstrap node becoming the leader just before being terminated