We need to deploy an upcoming Plone installation in Azure and have all freedom regarding the deployment. Moving everything into a VM would work. I am also thinking about using a container approach...what would you use in this case as Azure storage option for holding the Data.fs and blobstorage? The application has about 1 million cataloged objects, heavy on reads with a little requirements regarding wites...any real world experiences?
I started using K8s with Data.fs/Blobstorage. We do not have 1 Mio objects indexed, only about 200k. But this should not be an issue.
The key point on my end is read performance. It is generally awful if you use a block storage solution (usually network bound) or virtual storage like longhorn.io (usually CPU bound). I currently have both 
I will move to use direct local storage next year. Otherwise, I'm super happy about the K8s approach. With local storage, I will give up redundancy, but gain 10x read/write performance. The other thing is, you give up RWX volumes, which means in that context, that the cluster cannot "health" itself without any downtime by mounting Data.fs/Blobstorage to multiple containers.
I never deployed the cluster on Azure, though. Only on AWS, Google, Digital Ocean and Exoscale (European hosting)
With the help of AI cough, I managed to generate a Terraform configuration for a VM with local storage and a mounted blob storage using blobfuse2 for backups.