In designs like you describe (most of them out there), the plug is held in place from expanding wedges. When it's installed with good lubrication and torque it will sit there doing its thing indefinitely. It's solid and unmoving when the headset is adjusted or when the stem is changed.
In a design like you propose, the same bolt that adjusts preload on the headset also tightens the wedge in place at the same time. To get it started, you have the plug and topcap assembled outside of the fork completely (they have to stay together like this because the lower parts would just fall through if the bolt and topcap weren't there), and you engage the bolt enough that it fits snugly in the steerer as you install it. There needs to be a spring so that the wedge pieces are pressed down and engage when they're supposed to. When it's engaged just enough to catch, you push it down until it starts and begin tightening the bolt to wedge it in place while also drawing slack out of the headset.
Designs like this have existed. For a user or mechanic that's good at adjusting headsets and understands what the design is doing, it's not really a problem. But, a lot of people have difficulties with those things and don't need another analog or subjective feel-type element to master when adjusting their headset. I don't believe you can get away from needing the spring in these designs either, which implies steel, and while I don't have weights to compare, it probably makes it hard to save weight doing it that way.
One imagines that over-torquing the expander wouldn't be an issue, but a design like this in the wild in carbon steerers could prove otherwise. I could easily see a user switch stems and accidentally be left with insufficient stem gap. The plug you propose in that situation will no longer be splitting input force between preload and expansion. It will all go into expansion and the user won't know it, they just know that tightening the bolt takes out looseness. That could be a recipe for a destroyed steerer.