Cluster EVC Mode, Per VM EVC, and Hardware Versions

[BrettK1]

6 node cluster running 6.7u3, with EVC mode set to Ivy Bridge.
Looking to update EVC mode to Broadwell now that it no longer needs compatibility with an older separate management cluster, and soon may be able to set to Skylake, but that's at least a couple months out.

1) According to VMware documentation, Per VM EVC mode is disabled by default.  We have never enabled it on a per VM basis, and the templates that the VMs are created from those templates have Per VM EVC mode 'Enabled' and set to the current Cluster EVC level (Ivy Bridge).
a) Any idea WHY this could be occurring?
b) Does this mean that changing Cluster EVC mode will have no effect on these VMs even after a cold boot?
c) Is there any concern for VMs using older CPU EVCs than a Clusters higher setting, and therefore a need to make sure VMs are 'as current as possible' before changing a Cluster level?

2) There are many older VMs showing an EVC mode of 'Sandy Bridge'.  These VMs are (as far as spot checking has shown) older HW versions which do not have Per VM EVC available.
a) Can an OVA specify its EVC mode when it's an old HW version?
b) Are there HW/EVC compatibility issues (I haven't seen an article about this, so I'm assuming no, yet I've read off-site posts about issues with HWv8 and newer EVC modes, especially after Specter/Meltdown patches, etc).

I also saw a question here that wasn't answered about changing a cluster EVC mode failing due to some VMs being 'powered on', and this behavior being different than in 6.5.  As that was almost a year and a half old, wondering if there's any more info on this prior to even attempting it here.

[bluefirestorm]

Virtual hardware version can act as a CPU feature mask. For example, virtual HW version 11 or higher is required to have the Haswell CPU features (e.g. AVX2 instruction set) available in the VM. So if the virtualHW version < 11 while EVC is set at Broadwell or Skylake, the VM vCPU is as good as being stuck pre-Haswell. So it is probably a good idea to upgrade the virtual hardware version to the highest that ESXi 6.7 supports.

[BrettK1]

Thanks,

I'm less concerned about old VMs getting access to new features than just having it available for those going forward.  Most of our VMs that are pre-HW11 are because they are appliances that were deployed via OVA, and many either REQUIRE the older HW version (VCSA 6.7 itself for one example), or may experience issues tossing a new HW version at it (we tested a couple others, and some would generate much work for the Sysadmins).  Pretty much every other VM created from our own templates are on HW11 or higher.