fix: prevent close listener leak on rapid tunnel reconnections

[paulleseute]

Fixes the MaxListenersExceededWarning that shows up when tunnels reconnect a lot.

Every time a tunnel socket opens, we register a close handler on the Tunnel so we can clean up if someone calls tunnel.close(). That handler gets removed when the socket fires its close event.

The problem is when a socket dies and we immediately open a replacement — the old socket hasn't fired close yet (it's async), so its handler is still hanging around when the new one gets added. Do this enough times and handlers pile up until Node starts complaining.

Instead of waiting for the socket's close event to clean up, we do it right away when we know the socket is dead:

• TunnelCluster now passes the dead socket reference along with the dead event
• Tunnel keeps a map of socket → handler, and removes the entry as soon as dead fires
• Added an integration test that simulates 20 rapid reconnections and checks that listenerCount('close') stays flat

[punkpeye]

Just a minor lint issue

[paulleseute]

Just a minor lint issue

Fixed, PTAL

[fredericbarthelet]

I'm experiencing this too. Any chance this can get merged @punkpeye 🙏?

[github-actions]

🎉 This PR is included in version 1.4.1 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀

[punkpeye]

Thanks!