Daemon hot-loops on EADDRINUSE with no backoff or port-owner detection

[platypii]

Summary

When the gateway source cannot bind its port (EADDRINUSE), the daemon exits and the launchd/systemd service respawns it immediately, producing an unbounded crash loop with no backoff and no detection of who owns the port.

Evidence

Observed on a dev machine: 161 distinct daemon PIDs in ~30 minutes (one respawn roughly every 10s), each dying with:

listen EADDRINUSE: address already in use 127.0.0.1:8787

The port was held by an unrelated long-lived process. The loop ran for days, burning CPU, and was invisible at the CLI: every hyp command just printed a generic warning.

Problems

• No backoff or respawn cap; launchd KeepAlive + an always-failing daemon = hot loop.
• No detection/report of which process owns the port.
• No circuit breaker to park in a diagnosable stopped state.

Proposed

• Classify EADDRINUSE as a distinct, terminal-ish error class.
• Add respawn throttling (launchd ThrottleInterval / in-process exponential backoff) and a cap that parks the daemon in a clearly-stopped, diagnosable state after N fast failures.
• On bind failure, identify and report the owning PID/command.

Note

The user-facing error text for this case is being improved separately on branch hypaware-version (actionable EADDRINUSE message). This issue tracks the loop/backoff/detection behavior.