Really comes down to complexity of the setup. Like for us we end up piloting the upgrade 3-4 times just because of general testing. During that time we have a checklist developed of the go live tasks.
For go live we take a full DB backup, you’re setting up a new appserver anyways so you keep the old one. Worst case if it goes sideways, restore the DB and turn on the old appserver and you’re right back to where you started. Epicor has done a good job isolating versions these days.
Doesn’t take that much bravery LOL! I’m not so brazen as to even attempt in place Windows upgrades 
I would tend to disagree with this. It takes less time to upgrade your production app server than to do a DB restore and then Upgrade. We have our upgrade process down to a point I would even be willing to do it during production hours with a 1 hour window. We get all day Sunday, we’re 24/6, so it would be silly to do that, but if I had to I would do mid day.
There are a lot of variables that play into this including level of familiarity with the Epicor framework, size of your database, complexity of your customizations and integrations, IT support staff skill level, etc.
For a same major to same major I wouldn’t let them spend the $s on a new VM (i.e. 2023.1 to 2023.2) but if comfort level is higher then let them do it for 2023.x to 2024.x. Just my 2 cents 