Galaxy is built on top of kubernetes, so there’s no real direct correlation between the underlying server instances and the containers that run on top of it from an ecu standpoint.
You could look at it like this: They could be running clusters of m4.16xlarge instances, their resources are pooled in kubernetes. They then run containers with certain restrictions that resemble ecu’s. So a single 16xlarge instance could be running like 16 4-ECU galaxy containers for example.
In other words, they are not running containers directly on top of ec2 instances, but pooling multiple ec2 instances into a large resource pool on which they run containers. The underlying instance type doesn’t matter as long as it’s large enough that a single instance can support their largest container offering. In this case the minimum instance type would be xlarge.