Classically at least, the conservation of angular momentum has never been known to fail. As long as everything on the station stays on the station, the total angular momentum of the whole thing will remain constant. People can jump up and down, move around, make lots of heat, all go to one side, or the middle, it doesn't matter. The angular momentum of a closed system will remain constant.
The key there is closed system. If they vent gas or waste or shoot thrusters, then that can either add or remove angular momentum from the station (the make-up is in the angular momentum of the stuff moving away).
Sunlight or solar wind or other torques from the gravity of nearby objects can have very small effects that might build up over time, but people know about these things and the can be zeroed out by balancing torques.
So if you build a large rotating space station, or build it non-rotating and spin it up, it will pretty much rotate for millions of years unless you do something wrong. The only reason you'd need reaction wheels would be if you wanted to add attitude control, or adjust the speed of the spin slightly.
One reason that O'Neill cylinder come in counter-rotating pairs is to cancel out precession due to torque from other bodies. They need to remain pointed towards the Sun. One cylinder would keep rotating at about the same speed, but its axis could precess over time.
Another advantage is that the total angular momentum of the pair is zero. So you can build them at rest, and start spinning them in opposite directions using electric or other motors, without using any propellant.
So in some sense you could think of one cylinder as the reaction wheel for it's twin cylinder.
You could also use that with geometries other than cylinder pairs, run it as a net-zero angular momentum system with counter-rotating components. In that case you can have your reaction wheel if you want it.
Pair of O'Neill cylinders from here: