The longitudinal stress due to sustained loads is not constant - we often call it sustained stress but the Code does not. Let me call it SL here. Up until this point, we have been discussing SL based on support configuration; you get one number without liftoff and another number with liftoff. But SL doesn't "jump" from one value to another. It is a gradual transition based on the amount of thermal strain added to the system. If you could monitor the operating load on a +Y (vertical) restraint, you will see the load drop with no movement until the thermal strain up equals the deadweight down. At that point, the support then disengages. So what happens to the force-based (i.e. deadweight) load distribution as the system heats up? It changes. So SL changes.
It's not so black & white; SL changes with thermal strain. But looking at black & white will set the limits to SL and that's all we need.