Even in the case where f has low rank, I do not think it is given that you would want to run multiple invocations of f in parallel in e.g. f⍣≡. Even though f has low rank, ≡ does not, and so gets applied to the entirety of the result. Meaning that you need a synchronization after every iteration; if f is cheap enough, that may not be worthwhile. (And low-rank g in f⍣g is meaningless, as it needs to produce a single boolean.)

(You can do f⍣g⍤n, but then, again, that is rank doing the work of parallelization. And, performance considerations aside, I am no longer so convinced of the aesthetic utility of ⍤ as I once was. Wrt performance, when targeting CPUs, where branches are not so cheap, it might be worthwhile to design an f which can tolerate being called overly many times. Then give y an extraneous trailing axis the same size as a simd lane and do (f⍣g⍤1) y. That permits the whole computation to be vectorised, though it is also somewhat brittle.)

(Note ≡ is not the same as =. 1 1 1 ←→ 1 2 3 = 1 2 3, but 1 ←→ 1 2 3 ≡ 1 2 3. f⍣≡ is an idiom. f⍣= is meaningless except at rank 0, where is behaves the same as f⍣≡.)