The B-tree is not a competitor here; the additional processing has nothing to do with the search speed in the database, which is fast enough. This is a different issue. As for Kangaroo, its problem is that it is probabilistic, and as the search space expands, its success rate deteriorates. You know very well that adding DPS doesn’t work like magic.
Of course it's not magic. But what do you mean by this: "as the search space expands, its success rate deteriorates"? I'd say it's the other way around: having more pseudo-entropy increases the chances of a collision, because you have better uniformity, so the probabilities are allowed to behave closer and closer to the predicted "success rate". So this "success rate" as you call it is actually worse for smaller ranges, due to lower entropy.
Perhaps your DB has only a very specific use-case: brute-force, which, while guaranteed, you might never live enough for it to finish, even if you count billions of trillions of keys per second. In contrast, do you know of an example of a private key that Kangaroo cannot find? Because I never encountered one. It's like trying to pin-point to a single atom in the Universe and stating "hey, look, this one can never be found this way" and discarding the algorithm as problematic because of this. While all the other atoms are part of a single DP "tree" into which they all merge into, at some point.