I'm on team sets-as-fns and comp with the argument being that it most closely reflects my intent

Me too. It's easier to add to the set than add expressions to an or fn. It is also more mathematical.

I’m in the set of folks that prefer sets

I also like sets :) It's more concise, especially compared to (or (= ,,,) (= ,,,)) - Last time I wrote clojure with people who didn't have much clojure experience, I held a bit back on the idioms that I knew wouldn't be familiar. The or looks about how it would be in Javascript. I think it's fine to adapt coding style a bit to the prople one's working with - even if one solution is considered more idiomatic by experienced clojure programmers.

I also use sets, but I prefer naming the set and using an explicit get, e.g. (get valid-stuff x)

I've also done (valid-stuff? x) but stopped doing that after learning that predicates should explicitly return true/false (i.e. can't just be a set if you're a pedant like me).

Interesting - I haven't been naming my sets, I think that would be helpful.

Two or cases is ok… three or more starts to justify the idiom

How do people feel about an explicit contains? call: (contains? #{:a :b} foo)

I can go either way — sometimes it feels like it helps, sometimes it feels superfluous

I often add a (contains? ,,,), I find reading it a little easier. Might just be what I'm familiar with. I sometimes add a specific get for map lookup too.

Pointer equality vs. PersistentHashMap lookup It depends 🙃

For statically known sets, if the idiom repeats often enough, I usually write a macro that unrolls to an or of identity checks for up to quite a large number of keys

Is the goal of that macro improved performance / short circuiting?

Improved performance. A sequence of pointers comparisons will beat a set lookup probably up until the bytecode gets too large

Explicit 'contains' or '(or (= ...' are an easier way into Clojure for beginners/passing-byers, completely agree.

For statically known setsI have a macro in clj-kondo which creates a case expression out of this:

(one-of :foo [:foo :bar :baz])

Is case faster than (or (= ,,,) ,,,)?

it depends, benchmark and you will know, but usually (or (= ...)) is faster than set membership, which is why I often still use it if that's important

Case creates big ugly bytecode If the set is only keywords you can even use identical? Up to a ridiculous number of elements (or (identical? ...) ...) Will be faster and smaller than a case --- I also call that macro one-of 😄

It would take a very hot flame graph