The other thing the doc said that bothers me this:
“If a SAN is connected to the host, detach the fibre before continuing with the upgrade or migration. Do not disable HBA cards in the BIOS.”
Now this doesn’t make sense to me. How do people do in place upgrades on their cluster, because that’s production I can’t bring that down. My blades don’t connect to the SAN directly the Bladecenter chassis does.
This is more focused on an upgrade (installing from a previous version) as opposed to an update / patching (as stated in your post title). The risk is around a user or script accidently choosing the wrong LUN to install ESXi on and destroying production data. Typically a script chooses the first LUN it sees, so it is generally a good idea to mask off the other LUNs so that the host only has one choice. If you are just doing an update or patching, such as via Update Manager, there is typically no need to do this.
Note: In a blade environment, "detach the fibre" typically translates into putting a blade into maintenance mode and properly removing access to the production LUNs during an upgrade (unmount, mask, upgrade, remove mask, re-mount). As obviously you can't pull a cable 