Yes, that's a pragmatic solution. As long as the allowable preload is less than the max machining force, it's a neat and simple approach for ballscrew backlash.

It's easy to focus just on the ballscrew but I found on my machines that it's a stackup of several elements including the gibs (the table won't always move in the direction you want otherwise), spring and movement in the connecting elements (ballnut and ballscrew mounts etc) and of course the spindle itself. Unless your movements are friction free, there will also be some degree of springiness too. I managed to get the backlash in my HSK double ballscrews themselves down to under 10um but the total still ends up between 50-80um in some places.

You have to think about springiness as well as lost motion. Most Youtube Warriors who boast about "holding tenths" are measuring backlash with no forces acting, even if they are using a valid measurement technique. Once you apply machining loads, those figures will increase further.

The notion that you can simply dial out backlash in software is magical thinking. When you are climb machining manually, you will find there is a % engagement of the cutter where no significant force is required to feed the work into the cutter. Less engagement results in the work feeding itself, more requires you to push the work into the tool. So backlash will apply one way when you are using more engagement and the other way when using less engagement. The backlash model used in most controllers assumes conventional milling from what I've seen. The only solution is minimise backlash the very best you can. And only count on backlash compensation for conventional milling of shallow cuts (low % engagement) eg finishing passes. However, some cutters don't work well with light cuts, particularly when conventional cutting....