I never got around to digging into the literature when I first got curious about this question some years ago: is there any other Cilk-style fork-join scheduler that uses Rayon's StackJob trick where it can avoid per-job heap allocations entirely for 2-way fork-join? Since most uses of 2-way fork-join are for approximately balanced divide-and-conquer, it bounds the stack usage by the logarithm of the data size.