Properties of the servant program
Servant program as a worker server
contracts/common.md#principled-implementation A servant opens a WorkerServer that exposes one execute operation; the master decides counts and budgets, so the servant only needs the location and naming rules to find functions.
contracts/common.md#clear-and-simple-design A thin adapter over the worker server whose behavior lives in the private execute and iterate helpers.
Create a servant program.
Creates a servant program executing the prefixed functions in parallel.
Processing cost: Each execute request discovers matching modules once, then runs at most simultaneous loop workers and count invocations. Event storage grows with actual logger calls rather than an assumed one event per invocation.
Computation reuse: Discovery is refreshed for each execute request because the supplied filter and filesystem can change; requests themselves are effectful and cannot reuse earlier observations.
Resource ownership: The returned WorkerServer transfers connection lifecycle to its caller. Each execute owns its discovery collections, workers and event arrays until it completes; the master closes its servant process. No discovered-module cache is retained by this namespace beyond Node own loader.
Platform boundary: WorkerConnector process mode owns Node child-process launch/IPC and its explicit stdio modes. Feature discovery resolves the caller location with native path and fs realpath/stat, follows directory links with realpath cycle detection and imports discovered native absolute paths. It does not infer filesystem case policy from OS names.