When a node hits a node pressure eviction situation, the VM pods are killed, rather than Live Migrated. Customer expectation was that their VMs would live migrate, but instead were killed.
A node pressure eviction is not the same as an api based eviction like a node upgrade or administrator initiated eviction.
This unintentionally causes downtime for production workloads rather than smooth migration to other healthy nodes.

4.14 but applies to all versions

Force a node into a situation where we need to do a node pressure eviction:
https://kubernetes.io/docs/concepts/scheduling-eviction/node-pressure-eviction/

VM will live migrate (if eviction strategy is set to Live Migrate)