Because forking is so expensive (even if memory is copied lazily, the page table is copied eagerly, which makes it very costly when the forking process uses a lot of memory), this module is used to hold a number of processed forked off the main jenga process early on, when jenga uses little memory. Jenga can then ask these processes to cheaply spawn commands.
val init : Config.t ‑> args:string list ‑> unit Core.Or_error.t Async.Deferred.tmodule Reply : sig ... endmodule Request : sig ... endval run : Request.t ‑> Reply.t Async.Deferred.tval command : Async.Command.t