Probes are the kubelet's regular health checks on a container, and the three kinds act differently.
A readiness probe decides whether the Pod receives traffic. While it fails, the Pod stays Running but leaves every Service, and a rollout won't replace old Pods with it. A liveness probe decides whether the container is stuck: after failureThreshold failures in a row the kubelet kills and restarts it. A startup probe protects slow starters: until it passes, the other two wait.
Each probe uses one handler: httpGet (a path and port that must answer 200 to 399), tcpSocket (a port that accepts connections), exec (a command that exits 0) or grpc. Timing comes from initialDelaySeconds, periodSeconds and failureThreshold.
The classic mistake is a liveness probe that is too impatient for a slow start: the container is killed before it ever answers, over and over. The events show Liveness probe failed followed by Killing.
Ready for traffic, alive, and up to two minutes to start.