Update: I’ve made great progress on a continuation-passing implementation of effect handlers, with the temeritous hope that we can have both static guarantees that effects are handled, and performance comparable to OCaml. My concern with Hillerstrom’s work is that it doesn’t straightforwardly facilitate TCO - you have to pass a stack of continuations around in order to do so, which introduces a number of complications. But I think I’ve found a somewhat clever solution.