You are not overthinking.
1. We tried repeatedly to use Join Site, including getting assistance from Esri support. They could not make it work. I ended up using a "backup and restore" approach. We were upgrading from 10.8 to 10.9.1 and moving from a single machine to multiple machines (currently one for Portal and one for Server). It's interesting how well documented a process can be, can be recommended by Esri Support, and even Esri techs can't get it to work. My advice is 1. be flexible and ready to change directions and 2. just because it's documented doesn't mean it works.
2. Having done several upgrades now I can seriously say there is no easy way.
3. If you can manage services from Portal then your Server is federated, that's what that means. You should be able to see it in the Portal -> Organization -> Settings -> Servers page.
4. There is always some step left out, for me I struggled at several spots not knowing which of the myriad accounts and passwords I was supposed to use. You mean my account? Portal account? Server account? I do this kind of work off-hours, if you do, make sure you have every username and password available. Stupid password things are so fundamentally embarrassing to admit in public. 🙂 But can cost a day's work.
This probably won't happen to you but I keep saying this in Community because it was so painful. I think it was the 10.7 to 10.8 upgrade -- it crashed all every one of our WAB sites and left them unrecoverable (yes, that's what Esri told us after leaving us hanging on for weeks with "we're still looking at it" messages.) During the time they were "researching" we rebuilt all our sites in WABDE (WebAppBuilder, Developer Edition) which is free-standing so it does not depend Portal directly. This makes it more resilient. Each WABDE site has a separate copy of all the JavaScript it needs to function. By contrast the WAB sites all share code inside Portal, so they potentially all break at the same time on upgrades.
We keep our critical public site in Portal but have a version of it (always slightly out of date!) stored and available in ArcGIS Online. I think this is a good strategy for emergencies.
I wanted to migrate from a single machine to multiple machines for resilience, not performance. We were running an outdated SQL Server on the machine too, so we have a SQL Server managed by our county-wide IT group now, a Portal machine and a Server machine. There is a Datastore machine too, but the upgrade process would not switch over to it. By that point I threw up my hands and said "fine, Datastore, stay where you are" and will revisit it when I do the upgrade to 11.
One of the first things when there are problems that the Esri support tech says is "can you roll back to a snapshot?" and the answer used to be "no, that would wipe out our changes in SQL Server". Now that I have 3 servers instead of one, and a totally separate SQL Server, the answer will _sometimes_ be "yes, I can rollback".
Esri techs will sometimes ask you to do things like start an 8 hour upgrade process at 2 in the afternoon on a weekday. Sometimes they will ask you to roll back a server without telling you it might wipe out several days of work. You need to remember you are in charge and you need to ask lots of questions like "won't this take my business critical services off line possibly for the next 8-12 hours?" The techs can be very knowledgeable but sometimes they can also be a little too eager when you should be going slow. Make sure you know what is going on and if they can't tell you, stop and figure it out before you go on. You're not overthinking. 😉
