Guys celebi is not a fair example to use here. We've already agreed, u-turn lets landy get past celebi because, rey, as your damage calcs prove, the majority of the time u-turn is letting landy straight out beat celebi. Since the majority of celebis are slower than landy, celebi is simply not a check to u-turn landy.
OK, so if we actually want to talk about the use of u-turn, we need a counter where u-turn is not functioning like a coverage move, with latias being a good example. So:
If landy has u-turn, then here are the four possible scenarios (assuming for the moment that rachi uses u-turn):
Rachi stays in, landy EPs --> rachi KOed
Rachi stays in, landy u-turns --> you do some damage, take some damage, and lose momentum
Rachi switches, landy EPs --> you lose momentum
Rachi switches, landy u-turns --> opponent takes some damage and you gain momentum
So I do get what you're saying in that the options are overall "good" for the landorus player. But what would the situation be if u-turn didn't exist?
Rachi stays in, landy EPs --> rachi KOed
Rachi stays in, you double switch to ttar --> you take some damage (10% less) and lose momentum
Rachi switches, landy EPs --> you lose momentum
Rachi switches, you double switch to ttar --> you gain momentum
The only difference here is that landy does not do the ~20% from u-turn or take 10% from LO in the second scenario, and that you don't do any damage in the 4th. There's no crucial difference between the two scenarios! In fact, compare this to the 4 scenarios for something like terrakion vs an obvious switch (say xatu, since it also has u-turn for sake of example, and gliscor as the switch-in):
Xatu stays in, terra SEs --> xatu KOed
Xatu stays in, you double switch to LOmie --> you take some damage and lose momentum
Xatu switches, terra SEs --> you deal some damage, lose momentum
Xatu switches, you double switch to LOmie --> you gain momentum
Sound familiar? It's true that u-turn slightly alters the balance by allowing you to switch and deal a bit of damage in the same turn, but that's honestly not super relevant compared to KOs, as you pointed out in your r vs. r analysis, mike. I fail to see how u-turn substantially takes away the need to predict here (obviously, with celebi out of the picture).
And specifically @mikedawg, a couple problems there. First, celebi =/= other landy counters when we're talking about u-turn, since as I explained (and as rey's calcs support) when it's used vs celebi it's essentially a coverage move with the bonus of not having to predict at all. So celebi doesn't quite count here. And what you're describing is basically that when I have an offensive mon in vs something that I threaten, it's usually the best move to switch out...which I can take advantage of by predicting the switch and either u-turning or double switching. None of what you're saying is u-turn specific, and contrary to your assumption, oftentimes in high-level play people will go against r vs. r and do something like stay in on the u-turn (i've seen it happen in tournament matches), giving the rachi player just as signficant advantage as would the landorus played have had had the rachi player switched.