And if so, is regression more powerful if I build up my muscles from, say, the 90th percentile to the 99th percentile as compared to from the 30th percentile to the 50th percentile? As in, is the "regression" faster for higher levels of buildup?
Nearly all parts of our body will atrophy if we do not use them: muscles, ligaments, neurons. If you train to be able to deadlift 300 pounds then stop lifting, you will very quickly be unable to deadlift 300 pounds. You will eventually be unable to deadlift 250 pounds. A great while later, you might be unable to deadlift 200 pounds. Unless you are totally inactive--say, from paralysis or a coma--your atrophy will stabilize.
If you train hard enough to set a world record in, say, the clean & jerk, then stop training, your strength and power will rapidly fall from that peak. The regression to the mean will be rapid at first, since the last few percentile points of performance are the hardest to attain and maintain.
One major factor to consider is that long-held physical attributes are more resistant to atrophy than briefly-held ones. If you are able to lift 300 pounds (but no more) for several years, that ability will stay longer than if you were able to lift 300 pounds only after a period of training, then stopped as soon as you got there.