RequestHandlerRunner graceful shutdown limits under PHP‑FPM

Goal

Determine how far the built‑in graceful shutdown of laminas-httphandlerrunner's RequestHandlerRunner extends when the application runs behind PHP‑FPM rather than a long‑running async SAPI.

Constraints and uncertainty

• Signal‑based shutdown is opt‑in and requires the pcntl extension, which is often unavailable or restricted in shared hosting.
• In PHP‑FPM each request executes in an isolated worker; the runner’s signal handlers only apply to a persistent request loop, not to the per‑request lifecycle.
• Laminas does not provide an application‑level cancellation token or context, so propagating client disconnects or timeout signals into domain code is left to the developer.

Open questions

1. When PHP‑FPM sends SIGTERM to a worker, does RequestHandlerRunner invoke its registered signal handlers before the process exits?
2. Can application code reliably detect a client disconnect without attempting output, given that connection_aborted() only reports after a write?
3. Which configuration options (if any) control a shutdown timeout for in‑flight middleware when the runner is used in a long‑running SAPI?