That is only somewhat correct when taken literal. While it is more complicated (and fairly expensive) to track each and every individual migration (using schedtraces), you can definitely paint a picture, statistically speaking, by looking at esxtop. On a idle system with some vcpus at higher load, migrations (even intra LLC) might be fairly rare. You can look at MIG/s in esxtop (part of the "CPU Event Counts" fields).

There are reasons to be curious about this, not for capacity planning oe anything high level but esp. on AMD systems with multiple LLC domains per socket / NUMA node (depending on config) you might want to know where those vCPUs are when doing performance troubleshooting.

edit ~ two month later: just remembered that I updated a script that shows the placement and all affinity related information back then and forgot to add that. The caveat is that it is the vCPU/PCPU location at the time the vsi leaf was checked and depending on wakeup / switch / migrate rate, might not be very accurate: https://github.com/vbondzio/sowasvonunsupported/blob/master/vcpu_affinity_info.sh