I had a bug a little while ago and not sure if it's something that kondo could detect / warn

(defn any-evens?
  [coll]
  (reduce (fn [_ x]
            (when (even? x)
              (reduced true))) false coll))

(any-evens? [2 2])
=> true
(any-evens? [3 3 5])
=> nil

I wanted false in the second case

(defn any-evens?
  [coll]
  (reduce (fn [_ x]
            (if (even? x)
              (reduced true)
              false)) false coll))

gives the correct results ... I know nil is falsey but still

Not sure how kondo could help here (in general), as returning nil is valid in reduce functions.

(defn any-evens?
  [coll]
  (reduce (fn [_ x]
            (even? x)) false coll))

this is also correct, though not as efficient

Why not (some even? ....) ?

I'm trying to check whether using reduced and not having another part is flaggable

sure, it was a lot more complex in reality so I'm showing a trivial example ... I want to get to the fact that incorrect use of reduced ... forgetting to assign the result to the accumulator was actually the issue

maybe I need a better example