In our experience at polycloud, enabling Spot for EKS is one of the heaviest hitters when it comes to cost optimization - as long as your workload is compatible.

The interruption risk that is part of Spot can be mitigated by running on as many Spot pools as possible - each intersection of instance type and AZ is a separate spot pool, so it’s fairly easy to drastically lower your risk of interruption by deploying your Managed Node Groups across instance types and AZ’s.

For bonus points - build multiarch containers and add some Gravitron2 Spot pools!

cablespaghetti.dev gives us a nice background on the history of AWS Spot and the integration with EKS Managed Node Groups - great read!