I am evaluating a local diagnostic design before implementation or deployment and need to identify a public, supported macOS confinement mechanism. Proposed arrangement: A privileged custodian remains outside a separate privileged guardian's process group. The guardian launches a fixed diagnostic parent under a dedicated unprivileged identity. That parent sequentially launches three fixed sandboxed Python workloads, one child at a time. The current termination design targets the guardian's process group. It must not rely on whole-host process scans or indiscriminate killing. The proposed sandbox profiles are allow-default with file/network restrictions; the test-child profiles deny process-fork. We have not established that these restrictions prevent an existing process from changing its own group or session. Is there a public, supported interface or configuration that keeps all workload descendants within the supervisor's termination boundary from the initial child transition through final cleanup, while still allowing the parent to launch its authorized sequential children? In particular, please clarify: Escape through setsid/setpgid, spawn attributes, exec or native-library paths. When enforcement begins and whether descendants can relax it. Behavior when the diagnostic parent or guardian exits. Required privileges, entitlements, and supported OS/SDK versions. If the described sandbox categories do not establish that property, please identify the supported alternative boundary, if one exists. A different boundary would require an explicit design change on our side. I am requesting documented interface behavior and limitations—not private sandbox internals, a review of project code, or an absolute termination guarantee during kernel failure. No experiment has been performed to establish this property.