TLDR:
In the first introductory meetings I have with partners, I will admit to a bit of “salesmanship” saying over-simply that pods are running “directly on the ESXi kernel” along-side VM’s.  The problem is that neither Pods nor VM’s run “directly on the ESXi kernel.”

ESXi is not Linux… but containers must run on a linux kernel!?  The Pods are isolated into highly optimized and secure VM wrappers.  If you break out of the pod+container security, you’ll find yourself on a dedicated highly optimized and secured linux kernel, with no bios, no unwanted drivers, and no more ability to break out than any other VM.  All with a millisecond measured startup time suitable for container workloads.

If you trust VM security (which you must since you’re running vSphere), then you’re still gaining that security wrapper for container workloads.

Here’s the article that gives a good overview of how vSphere pod service pods are constructed and their inherent security:
https://blogs.vmware.com/vsphere/2020/05/vsphere-7-vsphere-pods-explained.html