Note that not all STM32 GPIO are 5V tolerant, you're needs to check datasheet for exactly your chip...
I know, they're pins with the "FT" IO level. I said "most" as I was acknowledging this but for all intents and purposes that doesn't matter as everything in question is an FT
So basically you use a 5V rail just to bias the pots when everything else is 3.3V? That's not very convenient and I'm not sure I'd get the idea if this is so. You said consistency, but what consistency exactly, please elaborate. Because if it's about wiper voltage vs position and you have to attenuate anyway, that's pretty moot. Anyway, as said above, you can't do it this way, it's going to load the pots leading to a completely different law. Just bias the pots with 3.3V, unless we have missed something major with your system.
No, everything else is 5v (external of the PCB), not 3.3v. Consistency meaning as part of a system everything in the system is 5v. It's easier to use 3.3v to feed the pots but the pots are not mounted to the PCB, they're mounted to a panel where the rest of the IO is 5v. While it's most likely to be reading pots, the analog channels may be reading some other analog thing and 5v is the "field voltage" or the supply to that analog thing.
To be honest I hate analog anything, I'm not super familiar with analog circuits. The analog input connectors are there just to be there, I usually use encoders. Since they're taking up PCB space they should work reasonably well. Never used an op amp so I was trying to avoid them but I suppose an op amp setup for unity gain avoids the "loading" issue? Most examples I see of using op amps to scale the input voltage use a resistor divider on the non-inverting input, so I assume that's going to have the same issue as just using a resistor divider by itself. I don't see too many examples of a resistor divider on the output of the op amp so thoughts?
I stuck the cap on the analog input here, I can't find any guidance on that so I took a guess