Moving to serverless is pretty straightforward if you can separate out your business logic into microservices. It's a mindshift for many development teams though, and as a result it become easier to keep thinking about "let me dump all this functionality into a container". For a lot of folks there is even a bridge to moving pure serverless by hosting images in Lambda - but again, its a mindshift to get to that point.

It's easier to start from scratch thinking about serverless, because as soon as you start down the containers path, it's REALLY difficult for teams to invest the time/energy/money into rearchitecting their platforms to support a pure serverless approach, even though it IS more efficient and it IS more fault tolerant than trying to build a virtual server farm of servers (which is what K8s and ECs becomes at scale).